Home

GeoDa™ 0.9 User's Guide - The University of North Carolina at

1. HIGH RISK ACTIVITIES The Software is not fault tolerant and is not designed manufactured or intended for use or resale as on line control equipment in hazardous environments requiring fail safe performance such as in the operation of nuclear facilities aircraft navigation or communication systems air traffic control direct life support machines or weapons systems in which the failure of the Software could lead directly to death personal injury or severe physical or environmental damage High Risk Activities The Copyright Holders specifically disclaim any express or implied warranty of fitness for High Risk Activities MISCELLANEOUS This Agreement represents the complet agreement concerning this License and may be amended only by a writing executed by both parties If any provision of this Agreement is held to be unenforceable such provision shall be reformed only to the extent 109 necessary to make it enforceable This Agreement shall be governed by the laws of the State of Illinois The application of the United Nations Convention of Contracts for the International Sale of Goods is xpressly excluded 110 Appendix B ANN License Agreement ANN Approximate Nearest Neighbors Version 0 1 Beta release Copyright 1997 1998 University of Maryland and Sunil Arya and David Mount All Rights Reserved This software and related documentation is part of
2. DISCLAIMER OF WARRANTY Software is provided on an AS IS basis without warranty of any kind including without limitation the warranties of merchantability fitness for a particular purpose and non infringement The entire risk as to the quality and performance of the Software is borne by you Should the Software prove defective you and not the copyright holder assume th ntire cost of any service and repair In addition the security mechanisms implemented in the Software have inherent limitations and you must determine that the Software sufficiently meets your requirements This disclaimer of warranty constitutes an essential part of the agreement SOME JURISDICTIONS DO NOT ALLOW EXCLUSIONS OF AN IMPLIED WARRANTY SO THIS DISCLAIMER MAY NOT APPLY TO YOU AND YOU MAY HAVE OTHER LEGAL RIGHTS THAT VARY BY JURISDICTION SCOPE OF GRANT You may use the Software on one or more computers use the Software on a network provided that each person accessing the Software through the network must have a copy licensed to that person copy the Software for archival purposes provided any copy must contain all of the original Software s proprietary notices You may not redistribute the Software in any form permit other individuals to use the Software except under the terms listed above modify translate revers ngineer decompile disassemble except to the extent applicable l
3. Figure 137 Moran Scatter Plot for Columbus CRIME Options gt Exclude Selected ON The Univariate Moran has several options invoked by using the options menu or by right clicking on the graph as illustrated in Figure 138 The Exclude Selected option works in the same fashion as for the standard Scatter Plot When the Exclude Selected option is active on the selection of observations in the graph or through linking the selection in any other graph will result in the recalculation of Moran s I for a layout without the selected observations The new regression line is shown in brown and the corresponding Moran s I is listed above the graph on the right hand side For example in NO Figure 139 the two observations highlighted in yellow on the left hand side of the graph have been selected The value of Moran s I on the right corresponding to the slope of the brown line is that for the dataset without the selected observations The slope of the blue line is for the complete data set LZ Exclude selected ON Randomization Envelope Slopes ON Save Results Save Image as Save Selected Obs Background Color Figure 138 Univariate Moran scatter plot options Moran colrook GAL CRIME Figure 139 Univariate Moran scatter plot with selected observations excluded While the observations are excluded from the calculation the spatial weights are not reconstructed but use a subset from the weight
4. Range Selection 30 lt CRIME lt 1100 Recoding REGIME ha Apply I OK Cancel Figure 96 Specifying a regime variable following a range selection NEIGNO PERIM REGIME 1005 000000 2 440629 0 1001 000000 2 236939 0 1 1006 000000 2 187547 1002 000000 1 427635 1007 000000 2 997133 E 1 1008 000000 2 335634 0 Figure 97 Regime variable added to table SR Save Selected Observations A new indicator variable with a value of 1 for the selected observations and 0 for the others can be added to the table by right clicking and choosing Save Selected Obs This brings up a dialog to select the variable name for the indicator variable as in Figure 98 The default is SELECT_1 but any other name can be specified Clicking ox will add a new column to the data table as illustrated in Figure 99 The new variable is immediately available for analysis but will not be added permanently until the table is saved as a new file Save Selected Observ EM Suggested Column Name Cancel Figure 98 Save selected observations in table NEIGNO PER IM 1005 000000 2 440629 1001 000000 2 236939 1006 000000 2 187547 1002 1 427635 REGIME SELECT_1 0 0 1 oco000 1 1 1007 000000 2 997133 1 1008 000000 2 335634 ha 1004 000000 2 554577 1003 000000 2 139524 1018 000000 3 169707 1 1010 000000 2 087235 Figure 99 Selected o
5. Color Save Image as Save Selected Obs Figure 35 Add Centroids to Table from Map EI Invoking this item brings up a dialog to specify the variable names for the coordinates Figure 36 Click on the check boxes and either choose the default or enter your own choice for variable names for the coordinates Click ox to add the new variables to the table In Figure 37 the last two columns of the new sidcent table are shown the original table is illustrated in Figure 33 with the coordinates added as new variables Note that the centroid coordinates only become permanently included in the data table after the table is saved to a shape file see the section on Editing and Manipulating Tables Add Centroids to Table Iw amp Coordinate xcoo z v Y Coordinate COO e Cancel Figure 36 Add Centroids to Table dialog m Table sidc XCOO LL 81 428840 36 372571 2 81 123626 36 434344 Sal 80 713185 36 337643 4 75 984813 36 401059 Figure 37 Columns with centroids added to the sidcent data table Creating a dBase File with Centroids A slightly different perspective is taken when the Add Centroids toolbar button is invoked or by using Tools gt Data Export gt Centroids This does not create a new shape file but only a data base file dBase format that contains a Key variable and the x and y coordinates but no other attributes This is useful when the coordinates need to be joined with a
6. KI 2 H Cluster Map The Cluster Map is selected by clicking on the matching check box in the dialog of Figure 152 The result is a special choropleth map showing those locations with a significant Local Moran statistic classified by type of spatial correlation bright red for the high high association bright blue for low low light blue for low high and light red for high low Figure 154 The high high and low low locations suggest clustering of similar values whereas the high low and low high locations indicate spatial outliers The Cluster Map can be linked and brushed in the same way as any other map in GeoDa 100 m 1 LISA Cluster Map colrook GAL _CRIME CA 1 LISA Cluster Map colrook GAL _GRIME Not Significant _ High High Low Low E Low High E hightow Figure 154 LISA cluster map for Columbus CRIME Box Plot The Box Plot is selected by clicking on the matching check box in the dialog of Figure 152 The result is a slightly customized box plot of the distribution of the Local Moran statistics as in Figure 155 This plot supports linking and brushing in the same way as the standard Box Plot Outliers identified in this graph pertain to the Local Moran statistics and not to the variable itself They are most useful as an informal diagnostic to identify pockets of local non stationarity Cressie 1993 Local Moran colrook GAL I CR
7. CONTIGUITY WEIGHT e Rook Contiguity The order of contiguity E SE C Queen Contiguity Figure 129 Specifying a higher order of contiguity Distance Band Spatial Weights When a shape file is specified as the Input File or when x Y coordinates are available as fields in a data table the spatial weights can be derived from the distance between these points This is carried out in the second panel of the Creating Weights dialog see Figure 130 The distance option requires input of the type of distance metric Euclidean Or Arc Distance the variable holding the x coordinate and the variable holding the y coordinate Note that these need not be actual coordinates but can be any two variables in the data table For a shape file the coordinates need not be made explicit The default is to use the x y coordinates in a point shape file and the polygon centroids in a polygon shape file It is important to use Euclidean Distance only for projected maps and Arc Distance for coordinates in unprojected latitude longitude note that the arc distance computed by GeoDa is approximate R3 CREATING WEIGHTS Input File shp C Program Files GeoDa Sample sidcent SH Ga Save output as C Program Files GeoD aS ample siddist1 quit Select an ID variable for the weights file CONTIGUITY WEIGHT s E ei DISTANCE WEIGHT Select distance metric lt Euclidean Distance gt v Variable for x coordinates lt Centroids gt X
8. GeoDa 0 9 User s Guide Luc Anselin Spatial Analysis Laboratory Department of Agricultural and Consumer Economics University of Illinois Urbana Champaign Urbana IL 61801 http sal agecon uiuc edu and Center for Spatially Integrated Social Science http www csiss org Revised June 15 2003 Copyright 2003 Luc Anselin All Rights Reserved Acknowledgments The development of the GeoDa software for geodata analysis and its antecedents has been supported in part by research projects funded by a variety of sources Most recent among these are U S National Science Foundation grant BCS 9978058 to the Center for Spatially Integrated Social Science CSISS Other funding was provided by NSF grant SBR 9410612 by a grant from the National Consortium on Violence Research NCOVR is supported under grant SBR 9513040 from NSF and by NSF grant SES 8810917 to the National Center for Geographic Information and Analysis Many thanks go to the students in the Fall 2002 class of ACE 492SA Spatial Analysis Department of Agricultural and Consumer Economics University of Illinois Urbana Champaign for being such good sports in testing an early version of the software Comments from users too many to list individually are greatly appreciated Yanqui Ren and Widodo Baroka provided research assistance to the development of earlier versions of the software GeoDa is a trademark of Luc Anselin All Rights Reserved GeoDa in
9. VE Cut off point Figure 131 Specifying the threshold distance for distance band spatial weights K Nearest Neighbor Spatial Weights Spatial weights based on a simple Distance Threshold criterion often lead to a very unbalanced connectedness structure For example this is the case when the spatial units have very different areas resulting in the smaller units having many neighbors while the larger ones may have very few or none yielding islands A commonly used alternative consists of considering the k Nearest Neighbors This is the second option in the DISTANCE WEIGHT section of the Creating Weights dialog K Nearest Neighbor weights are constructed by checking the appropriate radio button and specifying the order The default order is 4 but alternatives are readily specified in the dialog as shown in Figure 132 Again click on Create to start building the weights file and a progress bar will indicate successful completion Click on Done to close the dialog Treshold Distance Kl Cut off point e k Nearest Neighbors The number of neighbors ea Create Reset Cancel Figure 132 Specifying the order for k nearest neighbors spatial weights RS Characteristics of Spatial Weights The Tools gt Weights gt Properties command or the Weights Characteristics toolbar button creates a histogram of the frequency distribution of the number of neighbors in a spatial weights file The command brings up
10. a dummy variable can be saved to the data 45 table that contains values of 1 for the selected records and 0 for the others Choosing Options gt Save Selected Obs brings up a dialog in which the variable name for the indicator must be specified as in Figure 75 the default variable is SELEcT_1 Clicking ox adds a new column to the data table as in Figure 76 Note that this addition is not permanent until the table has been saved see the section on Editing and Manipulating Tables Save Selected Observ EM Suggested Column Name z Cancel Figure 75 Save selected observations indicator variable dialog SELECT 21 9 165903 13 929619 20 533881 22 140221 65 244668 71 902655 242 125984 174 698795 750 175932 745 330012 657 024793 673 014146 402 097902 397 142857 604 761905 624 579125 772 727273 709 243697 99 255583 86 359176 331 400966 476 456504 im Figure 76 Indicator variable added to data table Options gt Add Centroids to Table See the section on Manipulating Spatial Data A Smoothing Rate Maps Using choropleth maps to represent the spatial distribution of rates or proportions represents a number of challenges These are primarily due to the inherent variance instability unequal precision of rates as estimates for an underlying risk A number of procedures have been suggested to correct for this variance instability by smoothing the risk estimate GeoDa cont
11. 137 138 139 140 141 142 Change intervals dialog for histogram ssssesssesessssesessserresssereessssresssseressseree 65 Histogram for Columbus neighborhood housing values cceeesseeeeeeeteeees 66 e KEE EE 66 Box plot for Columbus neighborhood housing values using 1 5 and 3 as hinge 67 BOS ee Oe EE 67 Scatter plot of Columbus neighborhood crime on housing values 68 Scatter plot options EE 69 Selected locations on the Columbus neighborhood map using line select 71 Histogram with three highest categories selected eeeceeeeeeeeeeteeeeeteeeeeneeeeees 72 Box plot with selected Observations cccscccccssssseceeesnneeeeesseeeceeseneeesesenaseees 73 Scatter plot recomputed with selected observations exchuded 74 Linked box plot and box map sessssesssesessesssssrsseesssrtsseresseesseeesseessseesseessseessee 75 Select weight dialog to open an existing spatial weights Die 77 Adding a spatial weights file to the project 10 0 eeseeeseecesseeceeneeeeeeeeeeeeneeseeneeees 78 Using an already opened spatial weights file 78 Creating weights dialog visiscstnmcsiuanuiutsasidonsiue si dmasiiasan deans ashe 79 Output file specification for spatial We1g ht iin A ancient iwntionGdenteaaain 79 Kegy V aria ble specification EE 80 GAL format spatial weights file for North Carolina counties cseeeeeeeeees 81 GWT format spatial weights file for A order nearest neighbor
12. 232 18 801754 4 27 0 0 3 26 35 15 956 30 626781 3 89 1 0 4 33 200001 4 477 32 38776 3 7 0 0 5 23 225 11 252 50 73151 2 83 1 0 6 28 75 16 028999 26 066658 3 78 1 0 7 75 8 438 0 178269 2 74 0 0 Figure 53 Ascii text output file from Columbus data set polygon shape file 33 Both SpaceStat and ASCII options work in the same way After the menu item is invoked an Exporting Data dialog appears in which the Input file shape file and Output file name must be specified This operates in the same manner as in the Shape Conversion dialogs You need to click on the folder button to specify the Input file click open to confirm the choice and on the save button to specify the output file click Save As to confirm the choice The Exporting Data dialog then lists all the variables in the Input file that are available for export You select variables by clicking the double arrow gt gt to select all or individually by highlighting variables and clicking on the single arrow gt as in Figure 54 After the list of variables is complete click on Export to create the new file as in Figure 55 After the exported file is written to disk the Done button becomes active Click this button to close the dialog The new file will appear in the working directory EXPORTING DATA Input file name dbf Output file name C Program Files Ge Le COLUMBUS H a K BAM lt lt w Figure 54 Exporting data dialog E
13. C Queen Contiguity DISTANCE WEIGHT Select distance metric lt Euclidean Distance gt _ v Variable for x coordinates lt Centroids gt z Variable for y coordinates lt Y Centroids gt T Treshold Distance 1 Cut off point C k Nearest Neighbors The number of neighbors ja Create Reset Cancel Figure 127 Construction of spatial weights based on rook contiguity After selecting one of the two contiguity types click on Create to construct the weights A message will indicate successful completion Figure 128 Click on Done to remove the dialog R Input File shp C Program Files GeoD a Sample COLUMBU SI Save output as C Program Files GeoDa Sample colook gal NM Select an ID variable for the weights file POLY ID pA CONTIGUITY WEIGHT e Rook Contiguity The order of contiguity G S SHP gt GAL LIT TTT TTT TTT TTT Done Figure 128 Completion of weights construction Higher Order Contiguity Weights Spatial weights need not be limited to first order contiguity but higher order weights can be constructed as well In the creating Weights dialog change the order default value of 1 to any higher order Figure 129 The remainder of the procedure is identical to that for first order contiguity The higher order contiguity weights are based on the algorithm by Anselin and Smirnov 1996 that removes redundancies and circularities in the weights construction
14. Plot In Figure 143 this is shown for the Columbus crime data illustrating the degree of extremeness of the observed statistic 97 Moran colrook GAL CRIME AX Moran s I 0 5237 Figure 143 Envelope slopes in the Moran Scatter plot Options gt Save Results The Save Results option is invoked from the menu as Options gt Save Results or by right clicking in the Moran scatter plot window It allows you to add the standardized variable and its spatial lag to the data table for use in other analyses Selecting this option brings up a dialog Figure 144 to choose the variable to be added to the table by clicking on the corresponding check box and selecting a variable name The default variable names are STD_VAR and LAG VAR where VAR is the variable name used in the analysis Click ox to add the variables to the table as shown in Figure 145 As before the addition is not permanent until the table has been Saved As a different file Save Moran Plot Results wi Standardized Data STD_CRIME D Lag LAG_CRIME v Cancel Figure 144 Save results dialog in Moran scatter plot 93 STD_CRIME LAG_CRIME 1 159619 0 622430 0 975794 0 530835 0 269066 0 341683 0 163821 0 028828 0 932501 0 196220 0 541604 0 328404 Figure 145 Moran scatter plot variables added to the data table Other Univariate Moran Options As shown in Figure 138 the Moran Scatter plot also has three other common options Save Image
15. The Box Plot The Moran Scatter Plot Figure 152 Dialog for LISA windows Significance Map The Significance Map is selected by clicking on the matching check box in the dialog 99 of Figure 152 The result is a special choropleth map showing those locations with a significant Local Moran statistic as different shades of green depending on the significance level For example in Figure 153 the Significance Map is shown for the CRIME variable in the Columbus data set using rook contiguity The map you obtain may be slightly different since the results are derived from a randomization procedure that may yield slightly different significance levels depending on the number of replications used in the randomization procedure see also below Four significance levels are shown p lt 0 05 p lt 0 01 p lt 0 001 p lt 0 0001 As always in a randomization procedure the highest level of significance that can be obtained depends on the number of replications For example with only 99 replications the two more extreme significance levels will never appear The Significance Map can be linked and brushed in the same way as any other map in GeoDa 1 LISA Significance Map colrook GAL I_CRIME 9999 Permutation CJA 1 LISA Significance Map colrook GAL ILORIME Not Significant PS 0 05 EE oo E 0 001 E ooo A ih Figure 153 LISA significance map for Columbus CRIME geg S a
16. When the selection is carried out in the Histogram only whole categories can be highlighted However subsets of categories in a histogram may be highlighted as a result of selections in linked windows Note that in the current version of GeoDa you can only undo the selection in a Histogram by clicking outside the solid part in a Map A slightly more convoluted way to obtain the same result is to select all categories followed by Double Click to select the complement no categories Selection in a Box Plot In a Box Plot individual observations can be selected by clicking on the blue dots that correspond to them or by dragging a rectangle around them Shift Click will add more selected observations to the current selection In general Shift Click will change the selection status of individual observations The rectangle selection mechanism also allows you to select multiple observations 7 The selected observations are highlighted in yellow For values outside the interquartile range shown as dots below and above the main purple box in the Box Plot the selection is shown as a yellow dot for values inside the interquartile range the selection is a yellow line see Figure 116 for HovAL A Double Click selects the complement et BoxPlot Hinge 1 5 HOVAL R HOVAL Figure 116 Box plot with selected observations Note how in the current version of GeoDa you can only undo the selection in a B
17. Which representation is most useful depends on the analysis context For example for the calculation of spatial weights based on a common boundary a polygon representation is necessary In contrast when the spatial weights are derived from a distance criterion a point representation is needed The Tools gt Shape submenu contains the functionality to switch between point and polygon representations for the same data set and to import point coordinate information form ascii and dBase format data sets In addition it includes a function to export the boundary file information contained in a polygon shape file to several common ascii formats The Tools gt Data Export submenu allows selected variables from the current project to be exported in formats suitable for use as input into other statistical software Creating a Point Shape File with Centroids This function converts the contents of a polygon shape file to a matching point shape file where the points are the centroids of the polygons You have to specify the input shape file and a file name for the output shape file which will be a point shape file Specifically after selecting Tools gt Shape gt Polygons to Points a dialog opens in which the file name for the Input file polygon shape file and the output file point shape file must be specified The shape conversion dialog looks as in Figure 26 Click on the open folder icon to select the Input file simply typing in the name
18. or a national or resident of any such country or on any such list LIMITATION OF LIABILITY UNDER NO CIRCUMSTANCES AND UNDER NO LEGAL THEORY TORT CONTRACT OR OTHERWISE SHALL LUC ANSELIN THE REGENTS OF THE UNIVERSITY OF ILLINOIS OR THE DISTRIBUTOR BE LIABLE TO YOU OR ANY OTHER PERSON FOR ANY INDIRECT SPECIAL INCIDENTAL OR CONSEQUENTIAL DAMAGES OF ANY CHARACTER INCLUDING WITHOUT LIMITATION DAMAGES FO LOSS OF GOODWILL WORK STOPPAGE COMPUTER FAILURE OR MALFUNCTION O ANY AND ALL OTHER COMMERCIAL DAMAGES OR LOSSES IN NO EVENT WILL TH ISTRIBUTOR BE LIABLE FOR ANY DAMAGES IN EXCESS OF THE AMOUNT TH ISTRIBUTOR RECEIVED FROM YOU FOR A LICENSE TO THE SOFTWARE EVEN I HE DISTRIBUTOR SHALL HAVE BEEN INFORMED OF THE POSSIBILITY OF SUC AMAGES OR FOR ANY CLAIM BY ANY OTHER PARTY THIS LIMITATION O IABILITY SHALL NOT APPLY TO LIABILITY FOR DEATH OR PERSONAL INJURY TO HE EXTENT APPLICABLE LAW PROHIBITS SUCH LIMITATION FURTHERMORE SOME URISDICTIONS DO NOT ALLOW THE EXCLUSION OR LIMITATION OF INCIDENTAL OR CONSEQUENTIAL DAMAGES SO THIS LIMITATION AND EXCLUSION MAY NOT APPLY TO YOU Cl DI Kal e Wies Fi Sal Ia mi Dn Di D D om ls Fi J Cj
19. 2 while File gt Exit closes the program When there is an active project the menu contains four items Close to close the currently active window Close A11 to close all the windows in the project Export to save the currently active window as a bitmap file and Exit to end the program Two buttons on the Project Toolbar correspond to this functionality as well the open new project button see also Figure 1 and the close all windows button Project Toolbar Si Open project button ol Close all windows button Edit Menu View Tools Exe New Map Duplicate Map Add Layer Remove Layer Select Variable X Copy to Clipboard Figure 8 Edit menu The Edit menu contains three sets of functionalities as illustrated in Figure 8 The first set deals with manipulating maps the second set pertains to the selection of variables with the last related to the use of the Windows clipboard Edit gt New Map This item opens the file dialog to select a shape file The map will be added to a new window The new shape file should refer to the same spatial layout as in the existing map windows In other words after loading a map with the Columbus neighborhoods one could load a shape file with the Columbus Thiessen polygons that match these neighborhoods or with the neighborhood centroids a point file These three shape files all must have a matching Key Variable such as POLYID for which the unique values must pertain to the same location i
20. Color Save Selected Obs Save Image as T L Figure 111 Box Plot options A7 Explore gt Scatter Plot This function creates a bivariate Scatter Plot with the first specified variable Y on the vertical axis and the second variable X on the horizontal axis A least squares linear regression fit is superimposed on the scatter and its slope is listed at the top of the graph The window header lists the explanatory variable first X the dependent variable second Y as in the left pane of Figure 112 for the Columbus neighborhood variables CRIME and HOVAL Scatter Plot CRIME vs HOVAL _ O X Slope 0 6340 Scatter Plot CRIME vs HOVAL Ef Slope 05745 eames teat tant HOVAL teen at s anse 2 0888335 41 3420096 0 59518559 0 15163837 089946233 16452863 23921103 3 138934 CRIME Figure 112 Scatter plot of Columbus neighborhood crime on housing values The default is to compute the scatter plot for the data as they appear in the shape file Using the Options gt Standardized data command see Figure 113 the scatter plot can be shown for standardized values such that the slope of the regression line corresponds to the bivariate correlation coefficient GeoDa also shows the four quadrants of the scatter plot similar to the Moran Scatterplot so that it is straightforward to identify locations where above mean or below mean values on both variables coincide
21. ESDA Tool to Assess Local Instability in Spatial Association In M Fischer H Scholten and D Unwin eds Spatial Analytical Perspectives on GIS London Taylor and Francis pp 111 125 Anselin L 1998 Exploratory Spatial Data Analysis in a Geocomputational Environment In P Longley S Brooks B Macmillan and R McDonnell eds GeoComputation a Primer New York Wiley pp 77 94 Anselin L 1999 Interactive Techniques and Exploratory Spatial Data Analysis In P Longley M Goodchild D Maguire and D Rhind eds Geographical Information Systems Principles Techniques Management and Applications New York Wiley pp 251 264 Anselin L 2000 Computing Environments for Spatial Data Analysis Journal of Geographical Systems 2 201 225 Anselin L and S Bao 1997 Exploratory Spatial Data Analysis Linking SpaceStat and ArcView In M Fischer and A Getis eds Recent Developments in Spatial Analysis Berlin Springer Verlag pp 35 59 Anselin L R F Dodson and S Hudak 1993 Linking GIS and Spatial Data Analysis in Practice Geographical Systems 1 3 23 Anselin L and O Smirnov 1996 Efficient Algorithms for Constructing Proper Higher Order Spatial Lag Operators Journal of Regional Science 36 67 89 Anselin L and O Smirnov 1998 The DynESDA Extension for ArcView Bruton Center University of Texas at Dallas Richardson TX Anselin L I Syabri O Smirnov Y Ren 2001 Visualizing Spatial
22. Polygons The interface is similar to that for the Tools gt Shape gt Polygons to Points procedure see Figures 26 31 A Shape Conversion dialog is opened and you need to specify the Input file for example BALTIM SHP and the output file for example balthi shp The file names are specified by clicking on the folder button for the input file and entering the file name in the open file dialog followed by clicking open Similarly you need to click on the save file icon to select the output file followed by entering the file name in the Save As file dialog finish by clicking Save At this point the Shape Conversion dialog will show the name of the Input file a thumbnail for the point pattern associated with that file name and the name of the output file as in Figure 40 SHAPE CONVERSION Input file shp GI C Program Files GeoD aS ample BALTIM SHP Output file Tei lad balthi shp Create Reset Figure 40 Point to Polygon shape conversion dialog He Clicking on Create will start the computation process When the blue progress bar at the bottom of the dialog has reached the end a thumbnail for the Thiessen polygons will appear below the output file text box Figure 41 Click Done to end the procedure There will now be three new files in the working directory balthi shp balthi shx and balthi dbf The polygon shape file can be added to the current project by using Edit gt New Map Overlay the original points with
23. Ww a GI Figure 146 Bivariate Moran Scatter plot Moran Scatter Plot Matrix Univariate and bivariate Moran scatter plots may be computed for a number of variables and arranged as a scatter plot matrix as in Figure 147 for the Columbus variables CRIME and rnc On the bottom axis are the variables under consideration all in standardized units on the vertical axis the spatially lagged variables with the spatial lags applied to the standardized variables This provides an overview of the spatial patterning of each variable with itself as well as with the spatial lags of the other variables For example on the left hand side of Figure 147 the correlation between CRIME and neighboring CRIME is positive top panel whereas the correlation between CRIME and neighboring INC is negative Other combinations are insightful as well For example the relationship between Inc and W_CRIME top right panel in Figure 147 can be compared to the usual correlation between INC and CRIME left panel in Figure 148 as well as to the correlation between W_CRIME and Inc right hand panel in Figure 148 The latter is constructed as a regular scatter plot after the variables STD_CRIME and LAG_INC were added to the data table In contrast to the Univariate Moran it is legitimate to use the spatially lagged variable on the x axis in a Bivariate Moran since ordinary least squares remains an unbiased estimator for the slope of the regression
24. X Variable for y coordinates lt Y Centroids gt Treshold Distance 1 Cut off point k Nearest Neighbors The number of neighbors DEZ Figure 130 Distance weights creation In order to construct a spatial weights file based on a distance band the Threshold Distance button must be checked and a critical distance specified GeoDa internally computes the minimum distance required to assure that each observation has at least one neighbor This distance appears in the text box when the slider is clicked This is shown in Figure 131 for the SIDS centroids shape file sidcent shp and with Arc Distance selected as the method Moving the slider to the right increases the cut off distance The slider must be activated before a weights file can be created Click on Create to build the file and a progress bar appears indicating successful completion similar to Figure 128 Click on Done to close the dialog Note that the scale of the distance shown depends on the scale and projection for the coordinates used in the input shape file and may not be in any meaningful units Using the distance band criterion all points are selected that are within the specified distance from the observation under consideration R4 DISTANCE WEIGHT Select distance metric lt Arc Distance gt v Variable for x coordinates lt Centroids gt v Variable for y coordinates lt Centroids gt e t Treshold Distance 36 18267
25. a Weights Characteristics dialog in which the weights file must be specified Clicking on the open file button brings up the usual Open File dialog Spatial weights with either GAL or GwT file extensions may be specified as illustrated in Figure 133 Select the file and click on Open to complete the dialog then click on the ox button in the Weights Characteristics dialog to create the histogram WEIGHT CHARACTERISTICS Open weight file gal gut Cancel Look in jo Sample colgali GAL l colgal2 GAL colrook GAL File name Open Files of type Weight Files gal i Cancel Weight Files od Weight Files oa Figure 133 Selecting a weights file for the analysis of its characteristics The Connectivity histogram shows the number of observations on top of each bar by number of neighbors the numbers in the Histogram legend on the right You may need to change the default number of categories in the histogram to suit the distribution of contiguities use Options gt Intervals The Connectivity Histogram has all the properties of a standard Histogram and can be linked to the other views in a project For example in Figure 134 the North Carolina counties that have two neighbors are shown by selecting the right most histogram category Conversely by selecting one or more counties in the map as in Figure 135 one can find how many neighbors they have as defined in a gi
26. a default variable Once the Set the variables as default check box is checked there will no longer be a dialog needed to select the variable in subsequent analyses since all maps and statistical graphs will use the same default setting The menu item invokes a variable selection dialog as illustrated in Figure 12 The dialog lists two columns but only the left most Y will be used in univariate analyses The check box is marked to set the selected variable as the default If this check box is not marked the variable selection dialog will open for each mapping or statistical operation Variables Settings Select Variables 1st Variable Y COLUMBUS_ E POLYID NEIG HOVAL INC CRIME OPEN PLUMB v INC v Iw Set the variables as default R Cancel Figure 12 Variable selection Edit gt Copy to Clipboard Copies the contents of the current active window as a bitmap to the Windows Clipboard The contents of the Clipboard can then be pasted into any other software package such as a word processor or graphics package Alternatively any graph or map can also be saved to a bitmap file by using File gt Export or by right clicking on the active window and selecting Save Image As Edit Toolbar New map button Add layer button E Remove layer button 10 F Duplicate map button Zi 7 e Add centroids button see Tools gt Data Export gt Centroids he Select variable button Be Copy to Cl
27. a shape file that contains the data for your analysis The shape file can be either a point or a polygon coverage If your data is not in a shape file you need to convert it first either by using Tools for point data or by using ArcView ArcGIS or another GIS package that exports data as shape files Think of this shape file as the spatial reference for all your analyses You can have multiple geographical representations of the same data set for example as points irregular polygons or Thiessen polygons but they must all share a matching Key Variable see below GeoDa Project Setting Ed Input Map shp o gt Key Variable Figure 2 GeoDa Project Setting dialog In the Project Setting dialog a click on the open folder button will start the familiar Windows open file dialog Figure 3 Look in Sample EI ri Fav Stalttdes SHP BALTIM SHP COLUMBUS SHP SIDS SH Filename COLUMBUS SHP Files of type ESRI Shapefiles shp e Cancel Figure 3 Open File dialog Note that the files present in your working directory will likely differ from the ones shown in Figure 3 Using the standard Windows procedure you should navigate to the directory that contains the shape file you want to use in the analysis In Figure 3 that would be coLumBuS epp Click on Open to load the shape file Key Variable After you select the shape file you will need to specify a Key Variable This variable mu
28. a spatial weights files These three functions also correspond to a button on the Weights toolbar Details on their operation are given in the section on Creating and Manipulating Spatial Weights Weights Toolbar h Open Weights button WI Create Weights button Fz Weights characteristics button Tools gt Shape T Explore Map Options Window Weights VRlAlLl lERl Eidel eal Points to Polygons Data Export amp Polygons to Points Points from DBF Points from ASCII To Boundary BND Figure 16 Tools Shape submenu 12 The Tools gt Shape submenu Figure 16 contains the functionality to convert a point shape file to a Thiessen polygon shape file and to create a point shape file from the centroids of a polygon coverage Both input and output files in this operation are in shape file format In addition this menu contains functions to convert point data from either a dbf file or from an ascii file to a point shape file It also contains a feature to export the boundary file for a polygon shape to an ascii format including the data on the bounding boxes for the polygons Details are provided in the section on Manipulating Spatial Data There are currently no toolbar buttons corresponding to this functionality Tools gt Data Export Explore Map Options V Data Export gt SpaceStat ASCII Centroids Figure 17 Tools Data Export submenu The Tools gt Data Export submenu Figure 17 inc
29. out before interpreting the results of LISA maps as significant clusters or outliers The Randomization option provides a way to address numerical stability of the results The Significance Filter is designed to assess how conclusions depend on the chosen significance level and thus provides an informal mechanism to deal with multiple comparisons Randomization KC egies wale 0 05 Save Results KOEN 0 001 Selection Shape gt 0 0001 ERETT Figure 158 LISA significance filter option 103 1 LISA Significance Map colrook GAL Bislt m 1 LISA Cluster Map colrook GAL I_C EEX 1 LISA Significance 1 LISA Cluster Map Not Significar p 0 05 E 00 A Dee DW um Se DW Au Pat m H Ka et Figure 159 LISA maps after applying a significance filter Not Significa m High High ez AB Low Low E Low High E High Low eg RY Options gt Save Results The LISA Local Moran statistics for each location the association p value and the classification of significant at p lt 0 05 locations by type of spatial correlation can be saved to the data table by means of the Options gt Save Results command For both the Local Moran and the significance level the actual values are stored For the spatial correlation type an indicator value is stored which takes on the value of 1 for high high 2 for low low 3 for
30. the table as a new file to make the join permanent Join Tables EN Input file dbf C Program Files GeoDa Sample colky DBF Ki Key POLYID ad Do not include Include RECNUM gt _COORD Y_COORD Join N Cancel Figure 105 Join Tables dialog A3 NEIGNO PERIM RECNUM X_COORD Y_COORD 1005 000000 2 440629 1 8 795270 14 322100 1001 000000 2 236939 2 8 335270 13 986100 1006 000000 2 187547 3 9 027840 13 763000 1002 000000 1 427635 4 8 434140 13 714200 1007 000000 2 997133 53 9 084150 13 420700 1008 000000 2 335634 6 9865510 13 482700 1004 000000 2 554577 7 8 075840 13 327400 1003 000000 2 139524 8 8 374980 13 486000 1018 000000 3 169707 9 9615140 12 879900 Figure 106 Joined variables added to data table Undoing Changes Any changed made to the data table is temporary in that it only resides in memory In order to make the changes permanent the table must be saved as a new file To undo all changes select Refresh Data from the menu This will restore the table to its original state when it was first loaded from disk There is currently no more subtle undo facility All the changes are removed even those you may wish to keep The safest procedure to follow in this regard is to save the table each time you have added some new variables or made some edits that you are likely to use in the future 64 Statistical Graphs GeoDa contains standard EDA functionality in the form of
31. 000 2 000000 2 00000 3 165 000000 7 000000 1 000000 2 500000 1 000000 1 000000 0 000000 3 000000 2 000000 2 00000 4 104 300000 7 000000 1 000000 2 500000 1 000000 1 000000 1 O00000 2 000000 2 000000 2 00000 5 62 500000 7 000000 1 000000 1 500000 1 000000 1 000000 0 000000 2 000000 2 000000 0 O00000 6 70 000000 6 000000 1 000000 2 500000 1 000000 1 000000 0 000000 3 000000 3 000000 1 000000 7 127 500000 6 000000 1 000000 2 500000 1 000000 1 000000 1 000000 3 000000 1 000000 2 00000 8 53 000000 8 000000 1 000000 1 500000 1 000000 0 000000 0 000000 0 O00000 3 000000 0 co0000 9 64 500000 6 000000 1 000000 1 000000 1 000000 1 000000 1 000000 3 C00000 2 000000 0 000000 10 145 000000 7 000000 1 OO0000 2 500000 1 000000 1 000000 1 000000 3 O00000 2 D00000 2 0000 a1 A AAAA A ANNNAN 1 AANANAN I ANAAAAN A AAAAAN 1 ANAANANA A ANNAAAM I ANNAN I AAAAAA A AANA Figure 44 Baltimore point data as a comma delimited csv file After invoking Tools gt Shape gt Points from DBF a dialog appears in which the input file name and output file name need to be entered For example in Figure 45 the BALTIM dbf file from the Baltimore shape file trio is used and baltimpoint shp is specified as the output file You must click on the icon and enter the file name for both input and output files The x coord and y coord drop down lists show the variables con
32. 9193 14 6385 9 08714 14 6305 9 09997 14 2448 9 01505 14 2418 9 00895 13 9951 8 81814 14 0021 8 65331 14 0081 8 6429 14 0897 8 63259 14 1706 8 62583 14 2237 2 46 8 25279 14 2369 8 28276 14 2299 8 33071 14 2299 8 38366 14 2289 8 4446 14 2289 8 5445 14 2349 8 62413 14 237 8 62583 14 2237 8 63259 14 1706 8 6429 14 0897 8 65331 14 0081 M enn an MAAA E columbus1a bnd Notepad File Edit Format View Help 5597 14 7424 80945 14 7344 80841 14 6365 9193 14 6385 08714 14 6305 09997 14 2448 01505 14 2418 00895 13 9951 81814 14 0021 65331 14 0081 6429 14 0897 63259 14 1706 62583 14 2237 346 25279 14 2369 28276 14 2299 33071 14 2299 38366 14 2289 4446 14 2289 5445 14 2349 62413 14 237 62583 14 2237 63259 14 1706 6429 14 0897 Hesse ce Ee s Er agra Era Er see se tee ce Le ag Figure 49 Boundary file format 1 left and format 1a right E columbus2 bnd Notepad File Edit Format View Help E columbus2a bnd Notepad File Edit Format View Help 62413 14 237 5597 14 7424 80945 14 7344 80841 14 6365 9193 14 6385 08714 14 6305 09997 14 2448 01505 14 2418 00895 13 9951 81814 14 0021 65331 14 0081 6429 14 0897 63259 14 1706 62583 14 2237 62413 14 237 25279 14 2369 28276 14 2299 33071 14 2299 38366 14 2289 4446 14 2289 5445 14 2349 62413 14 237 62583 14 2237 63259 14 1706 stets b
33. ARSE ESEA 99 Bet 99 Significance Mapsen a E R A E E EEEa EE Ea E E E E 99 TUS ter EE 100 erte eege eege eeler 101 leet 102 LISA Option San aineina e E EA EE e 102 Options gt Randomization seseessseseesssesessssresssereesssereessseresssseressseree 102 Options gt Significance E 103 Options gt Saye results snisioritkosanii naeia ia 104 Bivariate E 105 E EE 105 IRGECTCHICES Eeer Ee Eege 106 Appendix A GeoDa License Agreement 108 Appendix B ANN License Aoereement 111 Appendix C Installing Grea E EE 112 Appendix D New in GeoDa 0 9 3 ec sessiesctacsscsviesvins aa esr eege esegg e 114 vi de Cl Ne E dE GA GA LA LA LA FA NO FA A A A LA FA WN N KK KK HK ke ke Fe Fe FL op E s z E ab Soe i kt ioh r S List of Figures GeoDa EE 3 GeoDa Project Setting dialog jasc wtviecssiav asseccevessec cx aa ics caag nev aas ates vaav ies 4 Openck Te E 4 Key Eeer gereegelt eher EE AER 5 Opening screen with selected shape file outline 0 0 0 eeeccceeeeseeeeeeeeteeeeeeeeaes 5 Moen anid toolbars eg arsina iaaiaee Aaa aA aN aaa 6 File Menu for new project and active project cccccccceeecceeseneeeeeeeeteeeeeeeeeeeeeees 6 Edit MEAN erste TA In EKDE iq Deg eegen See 7 Mu l uplenew EE 8 D pli ate maps EE 8 Point layer added to polygon layer with choropleth map for points 9 Varableselecthion EE 10 EE ee 11 Tools INEM E 11 Tools Weis Ee EE 12 Tools Shape SUDMECHU EE 12 Tools Data Expor
34. Autocorrelation with Dynamically Linked Windows In Computing Science and Statistics 33 Proceedings of Interface 01 Orange County CA June 13 16 2001 CD ROM Anselin L I Syabri and O Smirnov 2002a Visualizing Multivariate Spatial Correlation with Dynamically Linked Windows In L Anselin and S Rey Eds New Tools for Spatial Data Analysis Proceedings of a Workshop Center for Spatially Integrated Social Science University of California Santa Barbara May 2002 CD ROM Anselin L Y W Kim and I Syabri 2002b Web Based Spatial Analysis Tools for the Exploration of Spatial Outliers GIScience 2002 The Second International Conference on Geographic Information Science Boulder CO Sept 25 28 Assuncao R and E A Reis 1999 A new proposal to adjust Moran s I for population density Statistics in Medicine 18 2147 2161 Bailey T and A Gatrell 1995 Interactive Spatial Data Analysis London Longman Cressie N 1993 Statistics for Spatial Data New York Wiley 106 Fotheringham A S C Brunsdon and M Charlton 2000 Quantitative Geography Perspectives on Spatial Data Analysis London Sage Publications Marshall R J 1991 Mapping disease and mortality rates using Empirical Bayes estimators Applied Statistics 40 283 294 107 Appendix A GeoDa License Agreement Copyright 1998 2003 Luc Anselin and The Regents of the University of Illinois All Rights Reserved This software is subject to th
35. D aS ample colrook GA C Program Files GeoDa S ample colrook GAL C Program Files GeoD a S ample col Set as default Figure 136 Weights selection dialog Spatial autocorrelation analysis is implemented in its traditional univariate form as well as in a bivariate form Anselin et al 2002a Univariate Moran Scatter Plot The Univariate Moran Scatter Plot is invoked by Explore gt Univariate Moran or by clicking the Univariate Moran button on the Explore toolbar After the variable of interest and a spatial weights file are specified or using the default a window is created with a scatter plot that shows the spatial lag of the variable on the vertical axis and the original variable on the horizontal axis as shown in Figure 137 for the Columbus CRIME RR data using a rook based contiguity file The variables are standardized so that the units in the graph correspond to standard deviations The four quadrants in the graph provide a classification of four types of spatial autocorrelation high high upper right ow low lower left for positive spatial autocorrelation high low lower right and low high upper left for negative spatial autocorrelation The slope of the regression line is Moran s I listed at the top of the graph in blue The weights file used to compute the statistic is listed in parentheses colrook GAL in Figure 137 m Moran colrook GAL CRIME LI Moran s l 0 5237 H ce Oo
36. Edit gt Add Layer as in Figure 42 SHAPE CONVERSION Input file shp g C Program Files GeoD a S ample BALTIM SHP Output fie dei Richt Reset Done ancel Figure 42 TEE 7 i i point DEA with point data overlayed The data table for the Thiessen polygon shape file contains AREA and PERIMETER variables computed for the new polygons as well as all the attributes contained in the original point shape file as illustrated in Figure 43 Note that the AREA and PERIMETER 27 calculations are currently only supported for projected coordinates Euclidean distance For point shape files in unprojected Latitude and Longitude the results will not be correct Also the coordinates of the points themselves will not be part of this table by default but only if they had been added to the original point shape file explicitly by means of the Add Centroids to Table option Table balthi POLYID AREA PERIMETER STATION PRICE NROOM DWELL NBATH PATI 39 379200 23 625400 i 47 000000 4 000000 0 000000 1 000000 0 000 19 749900 17 556400 2 113 000000 7 000000 1 000000 2 500000 1 000 23 291400 25 572700 3 165 000000 7 000000 1 000000 2 500000 1 000 40 577900 24 947400 4 104 300000 7 000000 1 000000 2 500000 1 000 20 400000 18 465800 5 62 500000 7 000000 1 000000 1 500000 1 000 41 450700 26 483300 DG 70 000000 6 000000 1 000000 2 500000 1 000 Figure 43 Data table for Thiessen pol
37. ID 49 0 205964 2 199169 50 26 R49 ul 0 069762 1 102032 49 27 48 47 0 245249 2 079986 48 15 47 46 0 124728 1 841029 47 48 46 45 0 256431 2 193039 46 28 45 Figure 90 Records sorted by decreasing value of POLYID Editing Individual Table Cells Individual entries for table cells can be edited by double clicking on the cell The cell is highlighted and ready for editing as shown in Figure 91 Typing in a new value followed by Enter will change the entry in the selected cell The new value is used in all subsequent analyses but is not permanent until the table is saved as a different file Only then will the changes be written to disk Changes can be undone by selecting Refresh Data However this will undo all changes since the last save 55 m Table COLUMBUS AREA PERIMETER COLUMBUS_ COLUMBUS_I 0 309441 2 440629 2 0 259329 2 236939 3 0 192468 2 187547 0 083841 1 427635 Figure 91 Editing individual table cells N OV e a all Promoting Selected Records Records are selected by clicking on the left most column in the table as shown in Figure 92 Also clicking in the left most column and dragging down will select a set of consecutive records rows Records are also selected indirectly through linking and brushing of maps and statistical graphs see the section on Linking and Brushing The selected records can be moved up to the top of the table by choosing Promotion in the table menu The res
38. IME Mea ze CRIME Figure 155 Box plot for Local Moran statistics 101 Moran Scatter Plot The Moran Scatter Plot is selected by clicking on the matching check box in the dialog of Figure 152 The result is identical to the graph obtained with Explore gt Univariate Moran Figure 137 The scatter plot supports linking and brushing and all the same options as the standard Moran Scatter Plot If such a graph has already been created prior to the LISA analysis there is not much point in checking this option in the dialog LISA Options Each of the graphs and map windows created as part of a LISA analysis supports the same options as their standard counterpart They are invoked by using the options Menu or by right clicking in the graph In addition for the significance Map and Cluster Map three more options appear in the menu as shown in Figure 156 Randomization Significance Filter and Save Results The other map options are the familiar ones to change the look of the window save the map as an image and to create an indicator variable for selected observations R Randomization gt Significance Filter gt f Save Results Selection Shape d Zoom gt Color gt Save Image as Save Selected Obs Figure 156 LISA map options Options gt Randomization The Randomization option shown in Figure 157 allows the number of permutations to be specified and a new run of randomizations to be computed The
39. NRRAF AAA ANA nnanzi1i 8S j gt Figure 86 Computed rate variables added to data table 4 Variables Settings R TE R_EBS F_SPATRATE Figure 87 New rate variables available for analysis 53 Editing and Manipulating Tables The Table is included as one of the views on the data in the exploratory toolbox implemented in GeoDa The first time a request is made to analyze or map a variable in a map function or to construct a statistical graph the data table is loaded in memory and becomes available for editing and querying It can also be loaded explicitly using the Explore gt Table command or by clicking on the Table toolbar button Using the standard convention observations or records are shown as rows and variables or fields are shown as columns The Table functionality is invoked by right clicking anywhere in the active table This brings up a menu as shown in Figure 88 There are three main sets of functions The first set deals with selection of records the second with editing and data transformations and the third with saving and joining In addition sorting of records and editing of individual cells are invoked by means of mouse clicks x Promotion Clear Selection Range Selection Save Selected Obs Field Calculation Add Column Delete Column Refresh Data Join Tables Save to Shape File As Figure 88 Table menu Sorting Records by Field As initially listed the records in the table a
40. Open 2 x g Look in CQ Sample e Eier B Si colgali GAL Wsiddisti GWT Plsidr GAL Select from file gal gwt B colgal2 GAL Isiddist2 GWT sidrook GAL Si Blcolgueen GAL Plsiddist3 GwT sidtest GAL alcol 2 GAL lalsidk_6 GWT colrook GAL ed G Set as default Cancel File name Files of type Weight Files gal gwt Cancel Figure 119 Select weight dialog to open an existing spatial weights file 77 SELECT WEIGHT e Select from He gal gwt C Program Files GeoD a S ample sidrook GAL io Set as default S Figure 120 Adding a spatial weights file to the project SELECT WEIGHT e Select from currently used C Program Files GeoDa Samp C Select from file gal gwt Set as default L ee Figure 121 Using an already opened spatial weights file Creating New Weights New spatial weights are constructed by invoking Tools gt Weights gt Create or by clicking on the Create weights toolbar button This starts the Creating Weights dialog as in Figure 122 Before any weights can be constructed an Input File and Output File must be specified as well as a Key Variable The former is selected by clicking on the open file icon in the dialog and selecting a shape file select the file in the Open File dialog and click the open button The output file must be specified by clicking on the save icon and entering the file name in
41. Percentile SIDR79 Percentile SIDR79 E caw E 1 10 10 10 50 40 am 30 49 BE am mp E gt 99 2 Figure 68 Percentile map after zoom in Options gt Color The color options allow four different aspects of the map view to be customized the Map Shading Movie and Background Figure 69 Each of these operates in the same fashion Selecting an option brings up the Windows Color dialog as shown in Figure 59 Clicking on one of the color boxes changes the current setting for a map aspect to the new color The Color gt Map option alters the base color for the unclassified map i e the map that appears when selecting Edit gt Duplicate Map or clicking on the Duplicate Map toolbar button For example in Figure 70 this color has been changed from the default green as in Figure 5 to a red brown base The Color gt Shading option alters the color used for the hatch marks that identify selected observations see the section on Linking and Brushing For example in Figure 71 this has been changed to red from the default yellow The Color gt Movie option changes the color of the animated locations in the Map Movie Finally Color gt Background changes the color for the background in the map view For example in Figure 72 this is set to gray Note that the legend background color can also be changed after right clicking in its window pane You also use this short cut to save the legend pane to the clipb
42. US AREA PERIMETER COLUMBUS _ COLUMBUS I CONTIGUITY WEIGHT C Rook Contiguity The order d C Queen Contiguity DISTANCE WEIGHT Figure 124 Key Variable specification Spatial Weights File Formats The spatial weights are saved as text files with a Gat file extension for contiguity or a GwT file extension for distance They can be readily used by a number of other software packages as well The eat file format is illustrated in Figure 125 for the rook contiguity of North Carolina counties On the left the first part of a file is shown that results when the Key Variable is set to FIPSNO while on the right the file is illustrated for the case where no Key Variable was specified The main difference between the two files is in the first line the so called header line When a Key Variable is specified that line contains four values 0 reserved for future use the number of observations 100 the name of the shape file sIps and the variable name for the Key Variable FIPSNO When sequence numbers are used to label the observations the header line only contains the number of observations as in the right hand panel in Figure 125 After the first line the structure of the two cat files is identical For each observation there are two lines with information The first contains the observation Ip and the number of neighbors The second line contains the 1p values for the neighboring locations Note how in the lef
43. Vertical Tile Horizontal v 1SIDS Figure 23 Window menu 17 The Window menu Figure 23 lists all the open windows in Figure 23 this would only be the sIDs map and provides some means to rearrange the windows Cascade Tile Vertical Of Tile Horizontal Cascade stacks the windows Tile Vertical puts them side by side and Tile Horizontal arranges them vertically Note that in the current version of GeoDa this is the opposite of the usual convention The Linking on off switch is currently always on and cannot be changed it is grayed out in the menu There are no toolbar buttons corresponding to the Window menu Help Menu About GeoData Analysis Software Figure 24 Help menu The Help menu Figure 24 is currently not active and only contains the usual About item with copyright notices and credits Clicking on this yields an information box which lists the current version of GeoDa Figure 25 There is no built in help system in the current release of GeoDa About GeoData Analysis Sof EN Geoda 0 9 3 June 4 2003 developed by Luc Anselm amp Ibnu Syabri Copyright C 1998 2003 Luc Anselm and The Regents of the University of Illinois All Rights Reserved Figure 25 About GeoDa 1R Manipulating Spatial Data GeoDa is geared to the analysis of lattice data Observations are represented as spatial objects either as points with X Y coordinates or as polygons with the X Y coordinates of the boundary outline
44. XPORTING DATA Input file name dbf Output file name C Program Files Ge EI COLUMBUS Cl Export Figure 55 4 CRIME gt DISCED HOVAL Si POLYID Data ready for export 34 Mapping GeoDa contains rudimentary functionality for mapping and geovisualization This is limited to choropleth maps with some specialized flavors that aid in highlighting extreme values or outliers All maps are invoked from the Map menu Standard Choropleth Maps Getting Started The two most commonly used types of choropleth maps in GeoDa are the Quantile map and the Std Dev standard deviational map Before you can make a choropleth map you need to load the shape file with the data using the Edit gt New Map Edit gt Duplicate Map or similar commands In the window containing the layout for the shape file you can expose a legend pane by dragging the bar on the left hand side of the window slightly to the right as in Figure 56 dap Legend Figure 56 Exposing the Legend Pane You also need to specify a variable to be mapped If you earlier used the Select Variables toolbar button or Edit gt Select Variable you may have set a variable to be the default as in Figure 12 If that is the case you will not be queried for a variable name If you don t like the current default you will need to reset the selected variable by invoking Edit gt Select Variable and changing your choice If you have not specified a default variable t
45. ains five specialized mapping routines to deal with the visualization of rates in the Map gt Smooth submenu see Figure 21 Raw Rate Excess Risk Empirical Bayes Spatial Rate and Spatial Empirical Bayes Technical background and further illustration of these procedures is given in Anselin et al 2002b All five procedures require that a shape file be loaded and that two variables be specified one is the Event Variable the other the Base Variable The rates themselves need not be specified they are computed internally from the events counts of deaths disease incidence homicides etc and the population at risk the population to which the event pertains such as total births for the SIDS death rate The Event and Base Variable are specified in a Rate Smoothing dialog as in Figure 77 After selecting the variables click OK to generate the map RATE SMOOTHING X Select Variables Event Variable Base Variable Map Themes Cancel Figure 77 Rate smoothing dialog Figure 77 shows four of the five Map Themes available for mapping rates in a drop down list Percentile Map Quantile Map two Box Maps hinges 1 5 or 3 0 and a std 47 Deviational map These five types can be used for all but the Excess Rate map for which there is a customized legend The Spatial Rate and Spatial Empirical Bayes smoothing procedures also require that a spatial weights file be specified If there is no current default spatia
46. an also be individually specified for each category Double click on the square box in the legend and the standard MS Windows color dialog appears as in Figure 59 Select a new color from this palette by clicking on a square For example selecting a blue for the upper quintile will change the matching locations on the map as in Figure 60 Quantile CRIME 41st range 9 ahd range 10 ES 3rd range 10 4th range 10 Basic colors BR a range c10 BI FW WEE WI PE WPS EERE EE Custom colors III XI Define Custom Colors gt gt i Cancel Figure 59 Legend color customization dialog Quantile CRIME ox Quantile CRIME 1st range 9 e 2nd range 10 3rd range 10 P 4th range 10 i Sth range 10 Pei Figure 60 Customized legend colors gt ColorBrewer can be found at http www personal psu edu faculty c a cab38 ColorBrewerBeta html 37 A quantile map may yield surprising results when the data contain a lot of observations with similar values e g a value of zero for a map of rare events In that case some quantiles may be undefined since it is not possible to resolve ties and allocate observations with the same value to different groupings In such an instance all observations are given the color of the highest quantile for which there is no problem with ties For example if there were more than 10 observations with zero crime rate in the Columbus case there are 49 obse
47. ariable Computing Rate Variables Rates or proportions can be added to a table by using the Save Rates option in a rate map In addition they can also be calculated directly from the fourth tab in the Fiela Calculation dialog In addition to the five rate types implemented for the maps Raw Rate Excess Rate Empirical Bayes Spatial Rate and Spatial Empirical Bayes there is also the EB Standardized Rate which is used in the EB Moran Scatter Plot and LISA Maps see the sections on Global and Local Spatial Autocorrelation The operation is similar to the other dialogs specify the target variable name the method and a spatial weights file for the spatial rates and the Event and Base variables to compute the rates For example in Figure 104 the rate calculation is illustrated using the SID74 and BIR74 variables from the NC sips data set Field Calculation Unary Operations Binary Operations Lag Operations Rate Operations Result Methods NEWVAR e Raw Rate hu Weight Files Event Variables Base Variables SID 4 BIR74 Md Cancel Apply Figure 104 Rate calculation in a table A Saving and Joining Tables The changes and additions made to a table only reside in memory and are not permanent In order to make them permanent the table must be saved to a new file This is implemented in the Save to Shape File As command It invokes the usual Save As file dialog You must specify a file name which should b
48. as Save Selected Obs and Background Color These allow the scatter plot to be saved to a bitmap file to add an indicator variable for the selected observations to the data table and to change the background color for the graph For example in Figure 139 the background color was changed to gray These options work in the same way as described before Bivariate Moran Scatter Plot A bivariate Moran s I is invoked by Explore gt Multivariate Moran or by clicking on the Multivariate Moran toolbar button This creates a scatter plot with the spatial lag of the first variable on the vertical axis and the second variable on the horizontal axis Both variables are standardized internally such that their mean is zero and variance one and the spatial lag operation is applied to the standardized variables The slope of the regression line shows the degree of linear association between the variable on the horizontal axis and the values for the variable on the vertical axis at its neighboring locations as defined by a spatial weights file For example in Figure 146 the spatial lag of CRIME W_CRIME is on the vertical axis and the variable Inc on the horizontal axis Inference is carried out by invoking Options gt Randomization As for the Univariate Moran this computes a histogram for the reference distribution The Options are the same as for the Univariate Moran see Figure 138 and the discussion immediately afterwards 94 1 0
49. as in Figure 39 EXPORT CENTROIDS gt DBF Shape file shp C Program Files GeoD aS ample SIDS SHP E Output file dbf C Program Files GeoD a S ample sidcent_2 dbf LS Key Variable ARNM e ERR Figure 38 Export Centroids file dialog amp Microsoft Excel sidcent_2 DBF File Edit View Insert Format Tools Data Window Help Cea Gay A See 2 x 21h Ma Al v f RECNUM A B C D E 1 RECNUMIFIPSNO xX_COORD _COORD 2 1 37009 81 428000 36 372600 3 2 37005 81 123600 36 434300 4 3 37171 80 713200 36 337600 5 4 37053 75 984800 36 401100 6 5 37131 7 7 410000 36 372700 6 37091 76 993300 36 391300 8 H 37029 76 205000 36 379400 Figure 39 Centroids dbf file loaded into spreadsheet 25 Creating Thiessen Polygons as Shape Files Thiessen polygons are created as a polygon shape file derived from a point shape file Each Thiessen polygon encloses the original points in such a way that all the points in a polygon are closer to the enclosed point than to any other point This corresponds to the notion of geographic market area from economic geography Thiessen polygons are most useful for the visualization of variables when the point patterns are hard to distinguish In addition they allow the computation of contiguity based spatial weights for point data using the boundaries of the polygons to establish contiguity The Thiessen polygon procedure is started by Tools gt Shape gt Points to
50. as the original map As for Edit gt New Map this is the responsibility of the user since GeoDa assumes that it is the case but does not check for it The last layer added becomes the active layer on the map This is the layer to which selection linking and brushing are applied the lower layers are simply a backdrop This feature is useful when you want to add a point layer on top of a map showing areal units For example in Figure 11 the Columbus centroid points have been added on top of the original neighborhood layout and used to show a choropleth map for crime Note how the neighborhoods are not colored but only the points are The colors match the ones for the map on the right hand side of Figure 10 The current layer functionality is limited only the top layer is active and the order of the layers cannot be manipulated Quantile CRIME L ox Man Legend 1st range 9 hd range 10 3rd range 10 Ka 4th range 10 _ Sth range 10 Figure 11 Point layer added to polygon layer with choropleth map for points Edit gt Remove Layer This item removes the bottom layer in the active window When the last layer is removed the window is closed Currently there is a problem with this functionality in that layers are removed bottom up and not top down which is preferable Edit gt Select Variable Selects the variable s to be used for mapping and statistical analysis This is particularly useful when setting
51. avel aaa Ne R Ga aess 63 Lm e 64 EE 65 Explore KE EE 65 Lee EE 66 Explore gt Scatter let edd dees 68 Einkingand ETH 70 Selection ti a Mal wwe is aicivesiies aus otha vende Hawai Wan 70 Sclcetion in a MIStO EE 71 Selection in a box e EEN 72 Selection in a scatter plot EE 73 SElSCtON IN atable dagegen vecasd avsastes ancien eenegen 74 LE EE 75 CUS RENE eteegderee dergeint 75 Creating and Manipulating Spatial Weghts 77 Opening existing weights ees 77 Creating E 78 Spatial Weights file E EE 80 Contiguity based spatial weight 82 Higher order contiguity Weights iwdssises atitsasetsiined ais havitstines AE ENEE 83 Distance band spatial weights E 83 K nearest neighbor spatial weights a 5 assnceydenstiehsanslancszsadhateaviasdeastainselavuaeias 85 Characteristics of spatial Weights 3 c6scics tanceasesesicaaeswssascsaysaacedevieatetageannniveaies 86 Global Spatial Autocorrehktnem NENNEN ENNEN ee 88 Univariate Moran scatter Deet 88 Options gt Exclude Selected ON aen er clea ita EES 89 Options gt Randomization iesiccccicesieda cee DEER Nee E 91 Options gt Envelope Slopes ON 92 Options gt Save R s lts teste tune slaw ege 93 Other Univariate Moran options cccccceeescecceeeeneeeeeeseeeeeeseeneeeeeeeeaes 94 Bivariate Moran scatter Blots eeschte 94 Moran scatter plot MATIX are e aea EE A EE avin ae aaa 95 Moran scatter plot for EB rates ege ee eege deeg 97 Local Spatial Autocorrelation dese ri aaO AAE
52. aws specifically prohibit such restriction or create derivative works based on the Software rent lease grant a security interest in or otherwise transfer rights to the Software or 108 remove any proprietary notices or labels on the Software TITLE Title ownership rights and intellectual property rights in the Software shall remain in Luc Anselin and The Regents of the University of Illinois The Software is protected by copyright laws and treaties Title and related rights in the content accessed through the Software is the property of the applicable content owner and may be protected by applicable law This License gives you no rights to such content m TERMINATION The License will terminate automatically if you fail to comply with the limitations described herein On termination you must destroy all copies of the Software EXPORT CONTROLS None of the Software or underlying information or technology may be downloaded or otherwis xported or re exported i into or to a national or resident of Cuba Iraq Libya North Korea Iran Syria or any other country to which the U S has embargoed goods or ii to anyone on the U S Treasury Department s list of Specially Designated Nationals or the U S Commerce Department s Table of Denial Orders By downloading or using the Software you are agreeing to the foregoing and you are representing and warranting that you are not located in under the control of
53. box file is created as cover_r bnd where cover is the name of the polygon shape file The boundary file is created in the working directory and appears as in Figures 49 50 Exporting Shape to BND Format Type Input file shp C Program Files GeoD a ample COLUMBUS SHP g Type1 C Typela Output file text file C Program Files GeoDa S ample columbus1 bnd H C pel C Tore 2a Key AREA x Produce bounding box file Iv Yes coe Figure 52 Polygon shape to boundary file export dialog Exporting Data Tables The Tools gt Data Export submenu supports conversion of data tables between a shape file and output files in a format suitable for analysis by other software packages Specifically the spacestat option creates a binary file in the format used by SpaceStat The AscrII option creates a comma delimited text file csv format in the format suitable for input into many software packages It contains a header line that specifies the number of observations and the number of variables followed by a list of the variable names and a listing of the data by observation as shown in Figure 53 for a subset of variables from the Columbus data set The header line may have to be removed or edited to accommodate the requirements of specific software E columbus txt Notepad File Edit Format View Help POLYID HOVAL INC CRIME DISCBD EW cp 1 80 467003 19 531 15 72598 5 03 1 0 2 44 567001 21
54. bservations indicator variable added to data table Calculating New Variables Adding new variables to the data table is a two step process First a new column must be added and a variable name specified for the new field This is invoked by right clicking on the table and choosing Add Column This brings up a dialog to select the variable name as in Figure 100 Similarly Delete Column removes a field from the table 459 Add Column EM Input Column Name NEWVAR Ke Cancel Figure 100 Specifying a variable name for a new column field Once a variable field name has been created by adding a column its values can be computed using the Field Calculation command This brings up a dialog as in Figure 101 with four tabs each pertaining to a class of calculations unary Operations Binary Operations Lag Operations and Rate Operations Each tab brings up a slightly different dialog although the operation of the dialog is similar in each case you specify the target variable name Result the type of operation and any parameters needed for that operation such as a spatial weights matrix for spatial lag computations and the required variables variables The result is inserted into the new field As before this variable is immediately available for analysis but not permanent until the table has been saved as a new file In Figure 101 the dialog is shown for Unary Operations i e transformations of existing var
55. corporates licensed libraries from ESRI s MapObjects LT2 ESRI ArcView ArcGIS and MapObjects are trademarks of Environmental Systems Research Institute Other companies and products mentioned herein are trademarks or registered trademarks of their respective trademark owners GeoDa contains code derived from publicly available and licensed sources The Table functionality in GeoDa is derived from the MFC Grid Control 2 24 by Chris Mauder http www codeproject com miscctrl gridctrl asp The nearest neighbor calculations are compiled from source code contained in the ANN Code by David Mount and Sunil Arya see Appendix B for a copy of the License Agreement The Thiessen polygon code is based on source code by Yasuaki Oishi obtained from http www simplex t u tokyo ac jp oihsi src voronoi and described in Y Oishi and K Sugihara 1991 Numerically robust divide and conquer algorithm for constructing Voronoi diagrams in Japanese Transactions of the Information Processing Society of Japan 32 6 709 720 GeoDa Project Project Director Luc Anselin Software Design and Development Luc Anselin Ibnu Syabri and Oleg Smirnov Research Assistance Younghin Kho Yong Wook Kim Table of Contents ewe 1 Eroina totae a a aces cic stay es avis teas aes a ass Ait EE 3 Project s tting meninin nena ene aan be ee 3 Key i al AD fis elteren ee 4 DIST OV CLV IEW cee Cet EE 6 Grenier al E E 6 File meni Seano ain cease Shak ame
56. d Brushing Each of these functions is invoked in the same manner After selecting the menu item the Variable Settings dialog appears unless a variable or two variables for the scatter plots was earlier set as the default Edit gt Select Variable The spatial statistics Moran and LISA also require that a spatial weights file is selected Tools gt Weights gt Open Of Tools gt Weights gt Create After specifying these elements a window opens with the statistical graph or map The Options menu is context sensitive and contains ways to customize the graphs such as altering the number of categories in a Histogram or changing the Hinge for a Box Plot see the discussion of the Options Menu for each specific application Specifics for the various techniques are covered in the sections on Statistical Graphs Global Spatial Autocorrelation and Local Spatial Autocorrelation The Table item in the Explore menu opens a window with a table listing the variables as columns and the observations as rows In addition to an explicit invocation by means of the Explore gt Table menu or the Table toolbar button the data table is automatically 14 opened for any mapping or exploratory operation Detailed functionality is discussed in the section on Editing and Manipulating Tables Explore Toolbar Box Plot button ol Histogram button kl Scatterplot button Univariate Moran button l EB Moran button el Multivariate Moran button Je U
57. d R_SPATEBS for spatial empirical Bayes see Figure 85 for the raw rate example You don t have to accept the default but can specify any variable name in the text box Click ox to add the variable to the data table as illustrated in Figure 86 for various smoothed rates in the stps example The new rates are immediately available for analysis For example Figure 87 shows the variable selection dialog for a 41 box map of the raw rates without using the smoothed rates function As with all table manipulations the addition of rate variables is not permanent until after the table has been saved Choropleth Map gt Smooth gt Save Rates Add Centroids to Table Selection Shape gt Zoom gt Color gt Save mage as Save Selected Obs Figure 84 Save rates option for rate maps Save Rates BoxMap Hinge 1 5 Raw Rate SID79 w BIR79 Suggested column name K ON R_RAWRATE e K Cancel Figure 85 Rate variable name specification NWBIR79 R_RAWRATE R_SPATRATE 1364 000000 0 000000 19 000000 0 000000 0 001625 0 001485 542 000000 3 000000 12 000000 0 005535 0 002263 0 001730 3616 000000 6 000000 260 000000 0 001659 0 001862 0 001471 830 000000 2 000000 145 000000 0 002410 0 002030 0 002233 1606 000000 3 000000 1197 000000 0 001868 0 001956 0 002987 1838 000000 5 000000 1237 000000 0 002720 0 002148 0 002653 350 000000 2 000000 139 000000 0 005714 0 002177 0 002470 594 NNNNANN 2 NANNNANN 371 ANANAN NA A
58. data base file in a statistical package and a complete shape file is not required You can always load this dbf file back into GeoDa by converting it first to a point shape file using Tools gt Shape gt Points from DBF With the table functionality you can then use the Key variable to join the point shape file which will 24 not have any other attributes to another dBase format file with attributes However this will usually not be the way you want to proceed creating a centroid point shape file directly avoids the extra join operation and is more straightforward After clicking on the Add Centroids toolbar button or selecting Tools gt Data Export gt Centroids a file dialog opens as in Figure 38 You need to click on the folder buttons and enter the file names for both input shape file and output dbf files You also must select a Key Variable This will be included as a field in addition to the x and y coordinates so that the resulting dbf file can be easily joined with other data sets for the same observations An additional field contains a simple sequence number RECNUM which is used as a key in some statistical packages For the sips data the appropriate Key Variable is FIPSNO Clicking the Create button will then compute the centroids and write the output file You will note the new sidcent_2 dbf file in your working directory If you open it with a spreadsheet package like Excel you will see the four data columns
59. duce choropleth maps for variables that are expressed as rates Technical details are provided in the section on Smoothing Rate Maps 14 Map gt Map Movie Map Movie N Cumulative Geert Single Figure 22 Map Movie submenu The Map gt Map Movie submenu Figure 22 allows a choice between two ways in which the locations are highlighted in the movie Cumulative gradually fills out the complete map starting with the location with the lowest value for the selected variable and adding new highlighted locations as the values increase At the end of the movie the complete map is filled out with the highlight In contrast the Single option flashes each location in turn going from lowest to highest The speed by which this happens depends on machine hardware and is not predictable in the current version It is therefore recommended to use the Cumulative option More details on the Map Movie are provided in the section on Mapping Options Menu The Options menu controls several settings for specific statistical graphs and statistical analyses Only those options are shown in the menu list that are relevant for the graphs present in the program window at the time The look of the menu is therefore slightly differently in each case The Options menu items will be discussed with the techniques to which they refer The Options can also be invoked by right clicking on the active graph or map Window Menu Help Cascade Tile
60. e License Agreement set forth in the License Please read and agree to all terms before using this software 25 T BY INSTALLING THE GEODA SOFTWARE THE SOFTWARE YOU ARE CONSENTING TO BE BOUND BY THIS AGREEMENT IF YOU DO NOT AGREE TO ALL OF THE TERMS OF THIS AGREEMENT THEN DO NOT USE THE SOFTWARE Pl End User License Agreement The Regents of the University of Illinois grant you a non exclusive License to use the Software free of charge if a you are a student faculty member or staff member of an educational institution K 12 junior college college or university b you are a United States federal state or local government employee or c your use of the Software is exclusively at home for non business purposes Government contractors are not considered government employees for the purposes of this Agreement If you do not meet the requirements for free use of the Software you must contact the Spatial Analysis Laboratory at the Department of Agricultural and Consumer Economics of the University of Illinois Urbana Champaign the DISTRIBUTOR to obtain express written permission prior to any such use Commercial users must contact the Distributor to negotiate terms of use If you are using the Software free of charge under the terms of this Agreement you are not entitled to hard copy documentation support or telephone assistance
61. e different from the current shape file in the project This results in three files to be saved with file extensions shp shx and dbf The dbf file is the new table the other two files are copies of the shape file geography that was associated with the table If you only want to use the table as a data set in non GIS statistical software you can ignore or remove the shp and shx files Other tables can be joined with the current data table in a project provided that they have an exact match in terms of the number of records and the Key variable Right click on the table and choose Join Tables from the menu This brings up a Join Tables dialog In the example in Figure 105 the Input file colxy dbf is a file created by the Tools gt Data Export gt Centroids command and contains the variable POLYID as Key as well as a sequence number and the x_coorpD and y_coorpD centroid coordinates for the Columbus data set In the dialog you need to specify the Key variable in the drop down list and move the variables over that you wish to join with the current data table You use the same gt gt and gt symbols as in the Data Export interface e g Figure 54 Click on Join to carry out the procedure Three new columns will be added to the data table in your project as shown in Figure 106 Note that the Key variable is not duplicated The joined variables are immediately available for use in any of the analyses However as always you must save
62. e file However you can add the points to the table separately see next section on Adding Centroids to the Data Table Table sidcent Min PERIMETER CNTY_ NAME 1 0 114000 1 442000 1825 1825 Ashe gt 0 061000 1 231000 1827 1827 Alleghany 3 0 143000 1 630000 1828 1828 Surry 4 0 070000 2 968000 1831 1831 Currituck Figure 33 Data table for centroids point shape file 21 Adding Centroids to the Data Table In many instances it is not necessary to create a new shape file for the centroid points but the main interest is in adding x and y coordinates to the current data table These coordinates can then be used in distance computations to construct spatial weights In GeoDa this is accomplished as an option in any map view whether for polygons or for points When a map view is active the context sensitive Options menu contains Options gt Add Centroids to Table Figure 34 Alternatively right clicking on the map invokes an Options menu with the same items For example in Figure 35 this is shown for the sidcent point shape file Similar functionality is present for maps constructed from a polygon shape file Window Help Selection Shape gt Zoom gt Color gt Add Centroids to Table N Save Image as Save Selected Obs Figure 34 Add Centroids to Table option in Menu Choropleth Map Smooth Save Rates ee Add Centroids to Table a a B L Selection Shape gt zoom gt gt
63. eights file The EB Moran Scatter Plot appears as in Figure 150 Note that the actual variable names are not listed on the x and y axes but instead a generic RATES and W_RATES is used The variables names for the Event and Base are listed at the top of the window Note that the EB standardized rates in this analysis are not the same as the EB smoothed rates used in the map smoothing for technical details compare the Marshall and Assuncao and Reis references The EB Moran Scatter Plot has the same options as the standard Moran Scatter Plot They are invoked from the menu or by right clicking in the window 97 Moran s with EB Rate Standardization SID74 BIR74 Ef Moran s I 0 2714 Figure 150 Moran scatter plot for EB standardized rates The Save Results option allows you to add the standardized rate and its spatial lag to the data table The dialog is similar to that for the standard Moran scatter plot except that the default variable names are different As shown in Figure 151 the default for the Standardized Data iS STD RATES whereas the default for the Spatial Lag is LAG_RATES As always the addition of the new variables to the data table is not permanent until the table has been saved as a new file Save Moran Plot Results Iw Standardized Data STD_RATES D Iw Spatial Lag LAG_RATES E Cancel Figure 151 EB Moran save results variable specification dialog ON Local Spatial Autocorrelatio
64. ents over the expected number of events The expected number is computed by applying the average risk total number of events in all the locations over total population to the population at risk in each location Values of the Excess Risk less than one show locations with fewer than expected events values greater than one show locations where the number of events exceeds the expectation Note that the Excess Risk measure is a non spatial measure in that it ignores any effect of spatial autocorrelation The Excess Risk map uses a specialized legend that classifies locations by the extent to which they vary around 1 see Figure 80 for the 1979 SIDS data The map header lists the type of rate map as well as the events and base population used in the calculation Excess Risk Map SID79 over BIR79 Excess Risk Map SID79 over BI E um o 0 25 0 50 7 0 50 1 00 31 4 00 2 00 48 E zm 40045 E um Figure 80 Excess risk map for 1979 SIDS death rates in North Carolina counties Map gt Smooth gt Empirical Bayes This creates a choropleth map for rates that are shrunk towards the overall mean using the classic method of moments estimator in the Empirical Bayes procedure Marshall 1991 This smoothing procedure particularly affects the value for locations with small populations at risk the small area problem It will also typically remove the problem associated with many ties especially zero values in q
65. er details Explore gt Box Plot Creates a Box Plot for the specified variable The plot consists of a box going from the lowest value bottom to the highest value top The interquartile range is indicated by a A646 purple box in the middle of the graph its lowest edge is the first quartile 25 the upper edge the third quartile 75 A blue dot in this box corresponds to the median value The fence is indicated by a purple horizontal bar at the end of the T If the fence is below the lowest value or above the highest value the fence is drawn on the edge of the box In Figure 110 two Box Plots are shown for the Columbus Hovat variable with the hinge set to 1 5 default on the left five outliers and to 3 0 on the right no outliers All box plots are scaled to have the same size BoxPlot Hinge 1 5 HOVAL L Job BoxPlot Hinge 3 0 HOVAL Mee Figure 110 Box Plot for Columbus housing values using 1 5 and 3 as hinge The Hinge value for the Box Plot can be changed by means of the Options gt Hinge command The Options menu can also be invoked by right clicking on the plot window as in Figure 111 Besides the Hinge option which is unique to the Box Plot functionality the Background Color Save Selected Obs and Save Image as options work the same as for the other graphs and maps note how in Figure 110 the background color was changed to gray from the default white 15 Background
66. es of the polygon boundaries in a shape file to an ascii output file Four commonly used formats are supported allowing the input of these boundary files in a variety of mapping and graphics software packages Type 1 lists the polygon 1D the number of points in the polygon boundary followed by a list of x y coordinates as in the left panel of Figure 49 Type la right hand panel of Figure 49 is identical to Type 1 except that a header line is included with the number of polygons and the name of the variable used as Key Type 2 uses a common convention to cycle the coordinates in the list in the sense that for each polygon the last set of x y values is identical to the first Only the polygon ID is given not the number of points in the polygon boundary For example in the left panel of Figure 50 the values pair 8 62413 14 237 is listed twice for polygon 1 Type 2a is the same as Type 2 except for the inclusion of a header line with the number of polygons and the Key variable In addition to the boundary it is also possible to create an ascii file with the coordinates of the bounding box for each polygon This is used to derive contiguity information in the R spdep package The file contains an ID for each polygon and the x y coordinates for the lower left and upper right corners as in Figure 51 31 E columbus1 bnd Notepad File Edit Format View Help iL 14 8 62413 14 237 8 5597 14 7424 8 80945 14 7344 8 80841 14 6365 8
67. este ces Eeer s s ee Ar Era Erara Ee Kg 49 POLYID 62413 14 237 5597 14 7424 80945 14 7344 80841 14 6365 9193 14 6385 08714 14 6305 09997 14 2448 01505 14 2418 00895 13 9951 81814 14 0021 65331 14 0081 6429 14 0897 63259 14 1706 62583 14 2237 62413 14 237 25279 14 2369 28276 14 2299 33071 14 2299 38366 14 2289 4446 14 2289 5445 14 2349 62413 14 237 62583 14 2237 63259 14 1706 6429 14 0897 65331 14 0081 IDDDDAMDAMDMMDONADBDBDAMOOOODBDODAAE Figure 50 Boundary file format 2 left and format 2a right E columbus1_r bnd Notepad File Edit Format View Help 5597 13 9951 9 09997 14 7424 95009 13 7274 8 66655 14 2639 65331 13 5444 9 35149 14 0081 1986 13 5865 8 68527 13 8617 67758 12 8611 9 40138 13 7222 3333 13 2724 10 1806 13 6982 80197 12 942 8 45657 13 6445 10498 13 1041 8 73397 13 6444 12428 12 5952 10 0954 13 2985 10 10 0154 12 724 10 6497 13 2725 ONON e wN OO wN O o O Jo H H H H H H H H H Figure 51 Bounding box coordinates for Columbus polygons 39 After selecting the Tools gt Shape gt To Boundary BND command a dialog appears in which you need to specify the file names for the input a polygon shape file and output files an ascii text file as in Figure 52 In addition you must select one of the four output formats by checking the corresponding radio button If you also want the bounding box file you must check that option The bounding
68. g is most effective when multiple views are connected that pertain to different variables allowing a visualization and exploration of the multivariate association between variable In practice the number of views open at the same time is limited by screen size More importantly the simultaneous consideration of more than 9 11 graphs on a screen becomes a challenge Brushing Brushing is a dynamic implementation of the linking concept The selection in a given graph or map is continuously updated by moving a brush over the observations in a view In GeoDa the brush is a rectangle It is constructed by clicking and dragging the shape while holding down the ctru key After a few seconds the rectangle will start to blink indicating that brushing in now active The selection is updated when the brush is moved Brushing is turned off by a single Click anywhere in the view 75 Active brushing works for all graphs except the histogram as well as for multiple maps It is most effective between Scatter Plots and Maps especially when the Exclude Selected option is set Moving the brush over a Scatter Plot will immediately show how the slope is affected both in the form of a newly drawn regression line as well as by the changing slope value above the graph Brushing also affects the selected rows in a Table 76 Creating and Manipulating Spatial Weights Spatial weights are essential for the computation of spatial autocorrelation statis
69. gression slope m Scatter Plot INC vs CRIME EOX Slope 2 0407 1 1754 Figure 117 Scatter Plot recomputed with selected observations excluded Note that in the current version of GeoDa you can only undo the selection in a Scatter Plot by clicking outside the solid part in a Map A slightly more convoluted way to obtain the same result is to select all observations drag a rectangle around the whole Scatter Plot followed by Double Click to select the complement Selection in a Table Selection in a Table is carried out by clicking on the left side tab in the row corresponding to the selected observation Multiple rows can be selected by Shift Click on individual rows or by Click and Drag over the left side column Selected rows can be promoted as well as obtained as the result of range queries Details on these options are provided in the section of Manipulating and Editing Tables 74 Linking Any time more than one window is open in a project selection of observations in any of the windows automatically triggers the same selection in the other windows This is referred to as linking In Figure 118 a Box Plot for HOVAL is linked to a Box Map to illustrate how the five outlying observations are connected between the two views Boxhtap Hinge 1 5 HOWL Lower outlier 0 lt 25 13 25 50 12 50 75 12 gt 75 7 Upper outlier 5 Figure 118 Linked box plot and box map Linkin
70. h they require that a shape file has been loaded into the project In addition a variable must have been specified If no variable was set as default a Variables Settings dialog Figure 12 will appear to request a variable name Map gt Box Map A Box Map Anselin 1994 1999b is a special case of a quartile map where the outliers if present are shaded differently As a result there are six legend categories four base categories one for each quartile one for outliers in the first quartile extremely low values and one for outliers in the fourth quartile extremely high values The legend shows for each of the categories in parentheses the number of observations that fall in this category For the second and third quartile this is always _ of the number of observations For the first and fourth quartile this number will vary depending on how many outliers there are For example a Box Map of 1979 SIDS death rates in North Carolina counties reveals four upper outliers but no lower outliers as shown in Figure 62 This is invoked by the Map gt Box Map gt Hinge 1 5 the default To use a stricter definition of outlier select Map gt Box Map gt Hinge 3 0 BoxMap Hinge 1 5 SIDR79 Box lap Hinge 1 5 SIDR79 _ Lower outlier 0 lt 25 26 25 50 25 50 75 25 w e SEERE a t mm a i Figure 62 Box Map for 1979 SIDS death rate in North Carolina counties 20 Ma
71. he Variables Settings dialog will appear as in Figure 12 If you had previously specified two default variables bivariate variable selection the first one x will be used for mapping Note that the first time a data set is analyzed the data table is created and it may need to be moved out of the way to access the map functionality minimizing the table window will get it out of the way 35 Map gt Quantile In a quantile map the data are sorted and grouped in categories with equal numbers of observations or quantiles The Map gt Quantile command invokes a simple dialog to specify the number of quantiles or categories assuming a variable has been specified The default number of categories is 4 for a quartile map In the example in Figure 57 the number of categories was changed to 5 for a quintile map Clicking oK generates the map and associated legend as in Figure 58 for Columbus neighborhood crime Note how the legend pane indicates the type of map and variable name as well as in parentheses the number of observations in each category Quantile Map Eg of Classes Groups SG Cancel Figure 57 Quantile Map dialog Quantile Quantile CRIME 1st range 9 2nd range 10 D ara range 10 E 4h range 10 E A range 10 Figure 58 Columbus neighborhood crime quintile map He The legend colors are pre coded using the color schemes suggested in Cynthia Brewer s ColorBrewer color picker They c
72. he groups is given on top of the histogram bar In the current version the colors of the histogram bars are randomly assigned and cannot be changed nor can the break points be manipulated Intervals in the Hist EM of Intervals 12 Figure 107 Change intervals dialog for histogram AS m Histogram HOVAL Mele a skaed katres IECH 17 9 24 481667 IECH 24 481667 30 983334 EH EE H IL 37 525 48 066667 VC 44 066667 50 608334 a 50 608334 57 150001 E S57 15000 1 63 69 1666 63 69 1666 70 233335 E 70 233335 76 775001 Si 76 7500 1 83 3 16668 83 316668 89 858335 E 9 858335 96 400002 1 Figure 108 Histogram for Columbus neighborhood housing values Besides allowing a choice of the number of intervals the Histogram also has three other settings that can be customized These are invoked from the Options menu or by right clicking on the histogram window as shown in Figure 109 E Intervals Background Color Save Image as Save Selected Obs Figure 109 Histogram options The Background Color Save Image as and Save Selected Obs options work in the same manner as for the Map window see respectively Figure 72 Figures 73 74 and Figures 75 76 They allow the background color to be changed the window to be captured as a bitmap file and the selected observations see Linking and Brushing to be added to the data table as an indicator variable see the discussion in the section on Mapping for furth
73. he technology of dynamically linked windows Its origins trace back to initial efforts to develop a bridge between ESRI s ArcInfo GIS and the SpaceStat software package for spatial data analysis Anselin et al 1993 A second stage in the development of this idea consisted of a series of extensions to ESRI s ArcView 3 x GIS that implemented linked windows and brushing see Anselin and Bao 1997 Anselin and Smirnov 1998 Anselin 2000 In contrast to these extensions the current software is freestanding and does not require a specific GIS system GeoDa runs under any of the Microsoft Windows flavored operating systems Win95 98 2000 NT Me and Xp Its installation routine contains all required files GeoDa adheres to ESRI s shape file as the standard for storing spatial information It uses ESRI s MapObjects LT2 technology for spatial data access mapping and querying The analytical functionality consists of a set of C classes with associated methods A technical review of the design and architecture of the software is detailed in Anselin et al 2001 2002a Extensive background on the methodology of exploratory spatial data analysis linking and brushing and the specific techniques included in GeoDa and its predecessors can be found in Anselin 1994 1995 1996 1998 1999b This document serves both as a manual and as a brief tutorial for GeoDa It assumes some familiarity with basic GIS concepts as well as some knowledge of basic
74. iables Five such operations are currently supported equality Assign reverse sign Negate 1 value Invert Square Root and Log In Figure 101 the new variable NEWVAR becomes the square root of the variable CRIME in the Columbus data set Note that you can enter any numeric value instead of a variable name in the Variables text box Click ox to carry out the calculation Field Calculation Unary Operations Binary Operations Lag Operations Rate Operations Result Operators Variables NEWVAR v Assign vif CRIME xi Assign aA Negate NEWVAR CRIME mm Root bk Cancel Apply Figure 101 Field calculation unary operations AN Binary operations are mathematical operations on two variables or numerical values You invoke its interface by clicking on the second tab in the Field Calculation dialog As shown in Figure 102 five operations are currently supported Add Subtract Multiply Divide and Power You need to specify the Result variable and two existing variables Variable 1 and Variable 2 as well as the Operator For example in Figure 102 the variable NEWvAR becomes HOVAL AREA in the Columbus data set You could also replace either of the variables by a numeric value such as 1000 instead of AREA this would divide the house values by 1000 Click ox to carry out the computation Field Calculation Unary Operations Binary Operations Lag Operations Rate Operations Result Var
75. iables 1 Operators Variables 2 NEWVAR v JHOVAL X AR HOVAL AREA Figure 102 Field calculation binary operations Computing Spatially Lagged Variables Spatially lagged variables are weighted averages of the values for neighboring locations as specified by a spatial weights matrix This is implemented in the third tab of the Field Calculation dialog In addition to the target Result and variable you also need to specify a spatial weights file This file must first be opened or created with Tools gt Weights gt Open Of Tools gt Weights gt Create see the section on Creating and Manipulating Spatial Weights All currently opened spatial weights files will appear in the drop down list under Weights files if the list is empty you must open a weights file The spatial lag is implemented for row standardized spatial weights For example in Figure 103 a spatially lagged variable is constructed for CRIME in the Columbus data set using a rook based contiguity spatial weights matrix After the table is saved as a different file the spatial variables are available for use in other statistical software for example to include in spatial regression models in the R spdep package Al Field Calculation Unary Operations Binary Operations Lag Operations Rate Operations Result Weight files Variables NEWVAR v i C Program Files GeoDa Sample colrook e AREA v Figure 103 Computing a spatially lagged v
76. in North Carolina counties 40 Map gt Smooth gt Spatial Empirical Bayes Creates a smoothed map using the Empirical Bayes procedure but with the window average used as the reference for adjustment rather than the overall mean This results in some degree of spatial smoothing but less so than for the Spatial Rate smoother In Figure 83 this is shown for the 1979 SIDS data using the first order contiguity rook spatial weights to define the spatial window The map header lists the type of smoothing and the variables used in the calculation m BoxMap Hinge 1 5 SEBS Smoothed SID79 wi BIR79 Boxhfap Hinge 1 5 SEBS Srnoc IS Lower ouer 0 D lt 25 28 25 50 25 a ET 75 25 eae E E y an E Upper outlier 4 A Co K gt She Bag Sa Y Figure 83 Spatial EB rate box map for 1979 SIDS death rates in North Carolina counties Smoothing Options The Options menu for the smoothed rate maps is the same as for the other choropleth maps Specific to the rate calculations is the item to save the computed rates to the data table invoked by right clicking on the map and selecting Save Rates Figure 84 A dialog appears requesting a variable name for the rates Depending on the context a different default is suggested respectively R_RAWRATE for raw unsmoothed rates R_EXCESS for the relative risk R_EBS for the empirical Bayes smoothing R_SPATRATE for the spatial smoothing an
77. ipboard button View Menu View Tools Explore Map Project v Status Bar v Tools Figure 13 View menu The view menu Figure 13 contains two options to set which items are shown in the program interface and toolbar There are no buttons associated with these items View gt Toolbar Selects the toolbars that will be shown in the interface The default is that all four toolbars will be shown Project toolbar Tools toolbar Explore toolbar and Weights toolbar Unchecking one of these items will remove them from the window toolbar View gt Status Bar Sets viewing of status information in the status bar on or off The status bar is on the bottom and to the left of the program window The default is that status messages will be shown there Unchecking this item removes the status messages Tools Menu Explore t Weights gt Shape gt Data Export gt Figure 14 Tools menu 11 The Tools menu Figure 14 contains four submenus to deal with the construction and analysis of spatial weights the conversion and creation of point and polygon shape files and data export Tools gt Weights seg Explore Map Options W Open Shape gt Create H Ban Export Properties Figure 15 Tools Weights submenu The Tools gt Weights submenu Figure 15 contains three functions to Open a spatial weights file Create a spatial weights file and analyze the Properties of the connectedness structure in
78. ity definition was based ai colewt4 GwT E Of File Edit View Insert Format Help Oe lz a 49 COLUMBUS POLYID 06472 60118 22995 75306 SECH 60118 82845 3 6807 06472 18567 38497 97153 62214 0 1 1 1 1 2 2 2 2 3 3 3 3 D bh mt Go OD Ob DA Ga GG t t t t pwWwWNh ID db ww oO For Help press F1 Figure 126 GWT format spatial weights file for 4 order nearest neighbors in Columbus R1 Contiguity Based Spatial Weights Contiguity based spatial weights can be created when the input file is specified as a polygon shape file After both input and output files are specified both weights options in the Creating Weights dialog become active For the CONTUIGITY WEIGHT option a choice is available between Rook Contiguity and Queen Contiguity One of these radio buttons selects the type of contiguity criterion see Figure 127 for the use of the Rook Contiguity Rook Contiguity uses only common boundaries to define neighbors while Queen Contiguity includes all common points boundaries and vertices in the definition Spatial weights based on Queen Contiguity therefore always have a denser connectedness structure more neighbors CREATING WEIGHTS Input File shp C Program Files GeoD a S ample COLUMBU g Save output as C Program Files GeoDa Sample colrook gal RM Select an ID variable for the weights file POLYID v CONTIGUITY WEIGHT KI Contiguity The order of contiguity ES
79. l weights file in the project that is if it had not been previously set using a Tools gt Weights gt Open command and set as the default a warning will appear to set the weight You must first Open or Create the weights file You might also want to set it as default see Figure 78 SELECT WEIGHT e Select from file gal gwt C Program Files GeoD a Newsample rookgal GAL e je Set as default ls Create Cancel Figure 78 Specifying spatial weights for rate smoothing Map gt Smooth gt Raw Rate This computes the raw rate as a simple ratio of the event count to the base population at risk For example using the 1979 SIDS death rates for North Carolina counties this would require SID79 as the Event Variable and BIR79 as the Base Variable as in Figure 77 With the map type set to Box Map Hinges 1 5 the same map results as generated earlier in Figure 62 compare to Figure 79 The only difference is the map heading which specifies Raw Rate and the variables from which the rate is computed m BoxMap Hinge 1 5 Raw Rate SID79 wi BIR79 BoxMap Hinge 1 5 Raw Rate Lower outlier 0 lt 25 26 25 50 25 mm 75 25 WE 35 29 E Upper outlier 4 Figure 79 Raw rate box map for 1979 SIDS death rate in North Carolina Counties AR Map gt Smooth gt Excess Risk This computes a map showing the relative risk or excess risk as a ratio of the observed number of ev
80. line this is not the case in a regression of CRIME On W_CRIME 95 W_CRIHE Moran s l 0 5237 Moran s l 0 4712 W_CRIHE STD_CRIHE K IS 1 t n KE MAaEH UA 2 re w WM 1 LAG_INC Figure 148 Non spatial correlation matrices 96 Moran Scatter Plot for EB Rates The problem with variance instability for rates or proportions which served as the motivation for applying smoothing techniques to maps may also affect the inference for Moran s I test for spatial autocorrelation GeoDa implements the adjustment procedure of Assuncao and Reis 1999 which uses a variable transformation based on the Empirical Bayes principle This yields a new variable that has been adjusted for the potentially biasing effects of variance instability due to differences in the size of the underlying population at risk The EB standardized autocorrelation statistics is invoked as Explore gt Moran s I with EB rate or by clicking on the matching toolbar button This brings up a dialog similar in appearance to the one used in map smoothing the Event Variable and Base Variable must be specified as shown in Figure 149 for the sips data set RATE SMOOTHING EM Select Variables Event Variable Base Variable Set the variables as default Map Themes Cancel Figure 149 EB Moran s I variable selection dialog As for the standard Moran scatter plot you next need to specify a spatial w
81. llation problems Appendix D New in GeoDa 0 9 3 Fixes e Removed the 8 3 file name limitation e Fixed some issues with randomization procedures in LISA Streamlined Menu items and Options e Removed extra spaces in GAL file format eliminates a problem reading them into the R spdep package e Both GAL and GWT files with either sequence number ID or Key Variable New General Features e New Table functionality o Calculator new variables variable transformations standardization spatial lags rates rate smoothers o Queries sort select promote o Save edited table o Join table with other dbf tables with common key e Save selection dummy variable in all graphs maps and table new variable added to table e Save results of operations to data table rates smoothed rates spatial lags local Moran statistic classification p value e Window map customization for all graphs and maps background selection color etc e Newly created shape files centroid points Thiessen polygons include all variables from data file Statistical Functions EB Moran and EB LISA correction for variance instability for rates e LISA maps and significance maps have user specified number of permutations and significance filter for sensitivity analysis Input Output e Read point data from dbf ascii files not only from point shape files e Data export as comma delimited files csv e Capture maps and graphs as bitmap files e Export polygon bounda
82. lote cssaciinwanacnuiawae Wicca navtcaaanaces 95 Moran scatter plot MATIX sssini ech ea E ee anew Reale aes 96 Non spatial correlation matrices eenig Eege ege tele uierv eas 96 EB Moran s I variable selection dialog ccccccceeseccceeeeseeeeeeeeteeeeeseneeeeesenneeees 97 Moran scatter plot for EB standardized rates ce eeeseeeeceeeeeeeeeeeeeeeeeeeeeeneeeeees 98 EB Moran save results variable specification dialog ccccessecceeeesteeeeeeeteeees 98 Dialog for Ee 99 LISA significance map for Columbus cp 100 LISA cluster map for Columbus CRIME eueiett deisde rege 101 Box plot for Local Moran Statisttes suc ids aes ean Re eae 101 LISA map Options EE 102 LISA randomization option E 103 LISA significance filter option 103 LISA maps after applying a significance filter 0 0 eeeeeeeeeeeeeeeeeeeneeeeeneeees 104 LISA save results diglO Os as 50 vssetisaiasedereiasotictuaweneatiowtsatreiies aasaiasvieaslaes 104 LISA results added to data TEE 105 xi Introduction GeoDa is the latest incarnation of a collection of software tools designed to implement techniques for exploratory spatial data analysis ESDA on lattice data It is intended to provide a user friendly and graphical interface to methods of descriptive spatial data analysis such as autocorrelation statistics and indicators of spatial outliers The design of GeoDa consists of an interactive environment that combines maps with statistical graphics using t
83. low high and 4 for high low A dialog appears to select the variables to be added to the table and to specify the variable names as in Figure 160 The default names are I_VAR for the Local Moran cL_var for the spatial correlation category and PVAL_VAR for the p value where var is the name of the variable Clicking ox in the dialog adds the specified variables to the data table as shown in Figure 161 As always this addition is not permanent until the table has been saved under a new file name Note that in Figure 161 the observations have been sorted in increasing order of their p value with the observation with the most significant Local Moran listed at the top of the table Using selection sorting and promotion in the table allows you to investigate other significance ranges than the categories used by default Save Lisa Results IV Lisa Indices E CRIME Iw Clusters CLCRIME si CRIME Cancel Iv Significances PvAL_ CRIME x Figure 160 LISA save results dialog 104 LCRIME CL_CRIME PVAL_CR 2 037207 1 000000 0 000700 1 229924 1 000000 0 000800 0 212836 1 000000 0 001600 1 214552 2 000000 0 001800 1 289821 1 000000 0 005600 0 0 o 1 255139 2 000000 005900 1 217870 2 000000 007200 0 383639 1 000000 016900 Figure 161 LISA results added to data table Bivariate LISA The LISA principle can be applied to a bivariate measure of local spatial autocorrelation in a straightforward way It is in
84. ludes three ways in which to export data from a shape file to other formats The Sspacestat function converts selected variables from a shape file to the digital format used by SpaceStat the ascrz function creates a comma delimited text file csv in a format useful for input in a variety of statistical software packages and the Centroids function creates a dbf format file with the x and y coordinates of the polygon centroids Note that this is different from the Polygons to Points function above where the output file for the point coverage is in a shape file format this includes all three files with extensions shp shx and dbf and not only the dbf file Details are provided in the section on Manipulating Spatial Data The centroid export function has a matching tool bar button in the Edit toolbar Zi S SI Add centroids button 13 Explore Menu Map Options Wi Histogram Scatter Plot Box Plot Univariate Moran Multivariate Moran Moran s I with EB Rate Univariate LISA Multivariate LISA LISA with EB Rate Table Figure 18 Explore menu The Explore menu Figure 18 contains the functionality for traditional statistical graphics used in exploratory data analysis EDA as well as for spatial autocorrelation analysis In addition the Table is included as a additional data view in the form of a simple list All the graphs maps and tables are linked and all but the Histogram and Table can be brushed directly see Linking an
85. ms see Figure 6 Each of these deals with a particular functionality The most important menu items are matched by a button on the toolbar see Figure 6 The toolbar consists of four components that can be moved and docked independently anywhere within the program main window To dock a toolbar click and hold down the mouse on the raised vertical line to the left of the toolbar and drag it to the desired location The four components are the Project toolbar open a project close all project windows the Edit toolbar duplicating maps adding and removing layers choosing variables copying to clipboard the Explore toolbar EDA and ESDA techniques and the Weights toolbar creating and characterizing spatial weights A detailed overview of each of the menus and their corresponding toolbar buttons follows below File Edit View Tools Explore Map Options Window Help Lol aft acl Sef el E leee Figure 6 Menu and toolbars File Menu Edit View Too Close Ctrl C View Tools Close All Open Project Export gt Exit Exit Figure 7 File menu for new project and active project The File menu contains the standard Windows functions to open a project close the project windows and exit the program Figure 7 Its look is different depending on whether there are active windows open When there are no active windows in a project the menu only contains two items File gt Open Project invokes the GeoDa Project Setting Dialog Figure
86. n Local spatial autocorrelation analysis is based on the Local Moran LISA statistics Anselin 1995 This yields a measure of spatial autocorrelation for each individual location Both Univariate LISA as well as Multivariate LISA are included in GeoDa The latter is based on the same principle as the Bivariate Moran s I but is localized In addition the LISA can be computed for EB Standardized Rates The input needed for the LISA statistics is the same as for the global spatial autocorrelation statistics First one or two variable names or Event and Base for rates must be selected Next a spatial weights file must be specified for details see Global Spatial Autocorrelation Univariate LISA Univariate LISA Statistics are invoked by Explore gt Univariate LISA or by clicking on the Univariate LISA button on the Explore toolbar This brings up a dialog that lets you specify which of the four output options you want to generate as shown in Figure 152 The most relevant of these options are The Significance Map and The Cluster Map which are unique to the LISA functionality The Box Plot and The Moran Scatter Plot are identical to their standard counterparts with the only difference that the Box Plot pertains to the distribution of the Local Moran statistics Box Plot and Moran Scatter Plot also have the same options as their standard counterparts x Cancel What windows to open iv The Significance Map d he Cluster Map
87. n each of the shape files GeoDa currently does not check whether this is the case so it is possible to create nonsensical layouts Figure 9 illustrates this functionality Edit gt New Map has been invoked twice once for the colvor shape file Thiessen polygons and once for the colpoints shape file centroids Operations to create these new shape files from the original Columbus layout are illustrated below in the section on Manipulating Spatial Data In Figure 9 the three maps are shown side by side Window gt Tile Vertical O X COLUMBUS Figure 9 Multiple new maps Edit gt Duplicate Map Copies the layout in the active window to a new window This creates an exact duplicate of the original in terms of the boundaries or point locations but it does not retain the symbols or shading of the original map This is useful when mapping different variables to compare the spatial patterns across variables Figure 10 illustrates a duplicate map for the Columbus data with the first layout as a choropleth map for crime Note how the second window is only an outline the maps are shown after Window gt Tile Vertical COLUMBUS L IS quantile crime KOR Ast range 9 wd range 10 Ei 3rd range 10 E 4th range 10 E Sth range 10 Figure 10 Duplicate maps Edit gt Add Layer Adds a layer to the currently active map window The layer must conform to the same layout and the same Key Variable
88. n stare ege 6 Project TOG at ne aaa a ana a sepals a Aa AEA hatsmddevgalslatgadusceansies 7 Edit EE 7 Edit gt New Map ics e e E ck vista R i A oan EA a Ov as s 7 Edit Duplicate leet EE 8 Edit Add EE 9 Edit gt R move EE 9 Edit gt Select Variable snini niorir a RA E ett 10 Edit gt Copy Eege ey Sh awed accents evan eegen 10 Edit tOGI as oras einstan aa N EE EAEE AAAA Sege eeh 10 View EE 11 View TODA sss cay cass sity eds way cass a cay aac EAE aes a 11 AGW OLAS E 11 POTS MENU EE 11 TOO EEN aah ics te EE EE 12 KE Ee e oien enara a E A EE E E E AAA 12 Tools SHADE EE 12 Tools gt Data Ex POM EE 13 Explore meni EE 14 Explore 100 Bab egene ee 15 IVA APD TMG TALS Orgad es an EE 15 Map gt Box Map EE 16 IVA OO WEE 16 Map gt Map TEE 17 Options Men s EE 17 Keen EE 17 BAG Uo AUST oss os et cas cn EE Ee 18 Manipulating Spatial Data sdeteugeeeegeEe Eege 19 Creating a point shape file with eenegen sancecessudsisenkiomeuumuaanise 19 Adding centroids to the data table deggresieearetstaderereien deet g desda deed 23 Creating a dBase file with centroids ccccceecessccceeeeneeeeeseeneeeeeeeeneeeeeneeeeeeeees 24 Creating Thiessen polygons as shape files ccccceceesccceeeeteeeeeeeeneeeeeeetteeeeeees 26 Creating point shape files from dbf and ascii input Dies 28 Exporting bo ndary files aut ce caksecat as tinewts anata tacts Eaa E coast E E aE 31 Exporting data tables ee oh Ee 33 O OK 35 Sta
89. n to save the point shape file to the working directory As before the point shape file can now be opened in GeoDa Figure 48 illustrates how the point pattern and the data table are identical to the other incarnations of the Baltimore data Convert ASC to SHP Input file text file CAProgram Files GeoDa Sample baltim csy LG Output file shp C Program Files GeoDa Sample baltipoint2SHP gd coord P D Y coord STATION D Figure 47 Points from ASCH file and variable selection dialog an D sie DW Eg D a a D s Bo D a a E fa oo a 7 a a pe a s 7 D i de S af WEE CR a pn wf P S o 820 8 e KI D 3 I A D e D eng P p a a ob S S aa a D an a e D afp o D D Ba CS BR Ge Ze bd Ve EE We Bn oo SCH a 8 D Wi D D De e a a a e BD ew Se app 8 2 Be a oo D a gr d s An e SCH eg j HI a 5 e S Be NM i ple D DO _ STATION PRICE NROOM DWELL NBATH PATIO FIREPL 1 1 000000 47 000000 4 000000 0 000000 1 000000 0 000000 0 00000 2 2 000000 113 000000 7 000000 1 000000 2 500000 1 000000 1 00000 3 3 000000 165 000000 7 000000 1 000000 2 500000 1 000000 1 00000 4 4 000000 104 300000 7 000000 1 000000 2 500000 1 000000 1 00000 5 5 000000 62 500000 7 000000 1 000000 1 500000 1 000000 1 00000 Figure 48 Baltipoint2 point shape file created from ascii input Exporting Boundary Files The Tools gt Shape gt To Boundary BND command allows you to export the coordinat
90. ndard CHO Er E EE 35 Getting started erena e herean ana A a a AAA ARAE AAA uaa aA A AAR OAE nme 35 Map Quantes orseson e EE 36 O WER 38 Outlier ADS Ee Ee 39 Map gt Box Map EE 39 Maps EE 40 Map MOVi Cso aotar E EAER E EEE EEE E AA EREA 40 Map option EE 41 Options gt Selection EE 42 Options gt Zoomi EE 42 Options e 43 Options gt Save mage Hee D eg eech 45 Options gt Save Geelen Eeer geleet 45 Options gt Add Centroids to table eccececccceeeeneeeeeeeteeeeeeenneeeeeeeaaes 46 Smoothing Rate M ps cii oiii aa E E E E AA E E EA aad 47 Map gt Smooth gt Raw Kateinc cacti anita ante anesamarusim ese amicaia sin 48 Map Smooth gt EXCESS RISK enssins uain ntaa E A Aia A A Ria aaah 49 Map gt Smooth gt Ge RE 49 Map gt Smooth gt Spatial Kate eege geed Ae ee 50 Map gt Smooth gt Spatial Empirical Bayes ccsscccoserseossnctersnetorsnseersneeosats 51 KUER tee EE 51 Editing and Manipulating Table NEE ee Nee e 54 porting records by field si acsiacdnsasiunainomiuwadleshsiuwduaniuwasti 54 Editing individual table Cel Ga cc asc ce ec csv Galvan Sduscareth eege ege 55 Promoting EE EE EE 56 Clear Sele CHOU craked Eege ek 57 Range selecto EE 57 Save selected observatnonsg 59 Calculating jew variables sririsenrisssrosss sroin riisin ioi E aikai E RANNE aaa 59 Computing spatially lagged varables 61 Computing rate variables eege NENNEN 62 Saving aiid joining TABI ES nv cues hve dunes Nas Re AoA as G
91. ndition you can trick the upper bound inequality by adding a very small amount to the equality value e g to set CRIME 30 one could use 30 lt CRIME lt 30 001 Click Apply to generate the selection as shown in Figure 95 The selection is applied to all current graphs and maps in the project see the section on Linking and Brushing Range Selection Range Selection 30 lt CRIME z lt fio Apply Recoding peame Figure 94 Range selection interval specification 457 m Table COLUMBUS POLYID NEIG HOVAL INC CRIME 1 5 80 467003 19 531000 15 725980 2 1 44 567001 21 232000 18 801754 6 26 350000 15 956000 30 626781 33 200001 4 477000 32 387760 23 225000 11 252000 50 731510 6 8 28 750000 16 028999 26 066658 Figure 95 Range selection applied to table After the Apply button is clicked the second row of options becomes active under the heading Recoding Figure 96 This allows you to create an indicator variable that takes on the value of 1 for the selected records and 0 elsewhere The drop down list gives REGIME as the default variable name but this can be changed by typing in a different name The REGIME variable is added to the table as shown in Figure 97 The variable is available for analysis but the addition is not permanent until the table has been saved as a different file You can also skip this stage by clicking on ox which closes the dialog Range Selection X
92. nivariate LISA button Multivariate LISA button Ei EB LISA button E Table button Map Menu er V Quantile Percentile Box Map gt Std Dey Smooth gt Save Rates Map Movie gt Reset Figure 19 Map menu The Map menu Figure 19 contains the functionality to implement choropleth mapping The first four items correspond to standard map types Quantile map Percentile map Box Map and Std Dev standard deviational map Each of these requires that a variable be set Details are provided in the section on Mapping The Smooth and Map Movie submenus implement specialized mapping functions The Smooth group also includes an option to save the rates computed in the smoothing procedure Save Rates The Reset 14 item clears the map and returns the window to a simple outline There are no toolbar buttons matching the entries in the Map menu Map gt Box Map Options Window Help Quantile Percentile tt Droe 15 Std Dey Hinge 3 0 Figure 20 Map Box Map Submenu The Box Map item on the Map menu contains two options for the fences Hinge used to identify outliers Figure 20 These need to be specified before the box map is computed See the section on Mapping Map gt Smooth smooth o 7 ave Rates Excess Risk Map Movie gt Empirical Bayes Spatial Rate Reset Spatial Empirical Bayes Figure 21 Map Smooth submenu The Map gt Smooth submenu Figure 21 contains five methods to pro
93. number of permutations is applied to each observation in turn such that the total number of 109 computations equals the number of observations times the selected option For large data sets several thousand observations this can be time consuming Each new set of permutations yields a Significance Map and matching Cluster Map that take on a slightly different look This is a way to assess how sensitive the indication of significance is to the choice of the number of permutations 99 Permutations Significance Filter 199 Permutations Save Results 499 eS eae 999 Permutations Selection Shape gt Other R Figure 157 LISA randomization option Options gt Significance Filter The Significance Filter option shown in Figure 158 changes the threshold at which locations with significant Local Moran statistics are displayed on the Significance Map and the Cluster Map The default is p lt 0 05 which tends to yield more significant locations that would be warranted if an adjustment were made for the multiple comparisons involved in inference for local statistics of spatial correlation The Significance Filter allows you to set the threshold at a higher level of significance such as p lt 0 01 in Figure 158 As a result fewer locations will be portrayed as significant as illustrated in the LISA maps for the Columbus cRIME variable in Figure 159 It is strongly recommended that sensitivity analysis be carried
94. oard see Figure 72 Window Help Gelechon Shape gt Zoom gt Add Centroids to Table Shading Movie Save Image as Background i i Iel Figure 69 Map color selection options 43 Map Legend SIDS2 Figure 70 Changing the Map color oO KEE Figure 71 Customizing the Selection hash marks color m Percentile SIDR79 Percentile SIDR 9 lt 1 0 4 10 10 10 50 40 am 90 40 BB om am E 29 3 R Background color Copy legend to clipboard Figure 72 Customizing the Background color for a map view 44 Options gt Save Image as Besides capturing the currently active window to the clipboard Edit gt Copy to clipboard a map view can be saved to a bitmap file through options gt Save Image as This invokes a file Save as dialog in which the name for the bmp file must be specified in the usual fashion Figure 73 The saved file can then be incorporated into other documents For example Figure 74 shows the bitmap file that was saved by applying the Save Image as operation to the map in Figure 72 Save in O Newsample EI rt Ev File name sidsoreW I Save as type BMP format bmp EN Cancel p Figure 73 Save map as image dialog Figure 74 Saved bitmap image Options gt Save Selected Obs Whenever observations are selected whether explicitly on the map or through linking and brushing see Linking and Brushing
95. of the file will not work Select the input shape file for example stps in the open file dialog and click Open Figure 27 The file name will now appear in the Input file text box and a thumbnail of the polygon shape file will appear in the space below Figure 28 You select the output shape file using a similar procedure You need to click on the file save icon in the dialog simply typing in a file name will not work and enter the file name in the file name text box of the Save As file dialog Click Save to confirm the choice Figure 29 The file name of the output file will now appear in the text box of the Shape Conversion dialog Click on Create now active to carry out the conversion Figure 30 A thumbnail of the point shape file will appear in the space below the output file text box Once the progress bar at the bottom of the dialog blue line in Figure 31 is completely filled out click Done to leave the procedure 19 SHAPE CONVERSION x i SS TT SHAPE CONVERSION RE 2 x Look in Bsme S a e EM Elbalthies SHP BALTIM SHP OLUMBUS SHP Filename SIDS SHP Files of type ESRI Shapefiles shp EN Cancel D Figure 28 Shape conversion input file thumbnail 20 SHAPE CONVERSION X Input file shp I C Proaram Files GeoDa S ample SIDS SHP Output fie shp RI Save As Save in O Sample z e DI ek Ev lbalthies SHP EBALTIM SHP COLUMBUS SHP ISIDS SHP File name
96. ommand applied to the Columbus neighborhood crime data will show the outlying neighborhoods 40 first followed by a converging pattern to the center city This suggests a pattern of spatial heterogeneity where the lower values are found in the periphery and the higher values in the core A Map Movie requires that a variable is specified but no other dialog is used The Single option for the Map Movie is currently not very reliable and too dependent on specific machine hardware Map Options When a map is active the options Menu contains several ways to customize the map as illustrated in Figure 64 Similarly right clicking on any active map invokes a menu that includes several customization options In addition the right click also provides a short cut to the choropleth mapping functions discussed earlier in this section as well as to the rate maps see the section on Smoothing Rate Maps as shown in Figure 65 E Window Help Selection Shape gt Zoom gt Color gt Add Centroids to Table Save Image as Save Selected Obs Figure 64 Map Options menu Choropleth Map gt Smooth gt Save Rates Add Centroids to Table i Selection Shape gt zoom gt Color gt Save Image as Save Selected Obs Figure 65 Map Options and shortcuts context menu Al Options gt Selection Shape The Selection Shape is the graphical mechanism by which observations are selected on a map GeoDa supports five diffe
97. operating system C WINNT SYSTEM32 C WINDOWS SYSTEM32 Or C WINDOWS SYSTEM Until you have access to a stable release of GeoDa it is highly recommended that you stick with the default location for the files files in bold must be registered on the target computer the installation program should take care of this but you can always do it manually as well refer to the Windows Registry instructions for your system mfc42 d1l1 Msvcirt dl Msvcp60 dl msvert dll Msvert40 dl1l oleaut32 d1l olepro32 d11 Stdole2 tlb GE Installation Error Messages Sometimes MS Windows reports an error during installation such as Msvcirt dll is linked to missing export msvcrt dll The likely cause is that GeoDa s installation routine cannot find the correct version of the msveirt dll file Another program may have installed a different version of the file or the file may have been damaged It is important to note that there are two files with very similar file names that can easily be confused Msvcirt d1l1 and msvert d11 The latter is used by the MS operating system and should never be replaced or renamed To fix the problem e Search the computer to locate all copies of msvcirt dll e Rename each of these files e g to msvcirt old e Reinstall GeoDa e Reboot the computer before running GeoDa Check the Openspace mailing list http sal agecon uiuc edu mailman listinfo openspace for any other messages pertaining to insta
98. or alternatively locations where above mean or below mean values in one variable coincide with below mean or above mean values for the other This is illustrated in right hand pane of Figure 112 As is the case in all the graphs the Options menu for the Scatter Plot can be invoked by right clicking on the graph as illustrated in Figure 113 Besides the Standardized data feature an important option is Exclude selected Choosing this activates the dynamic recalculation of the regression slope as observations are selected The new slope AR excludes the selected observations Typically this is used when brushing a scatter plot or map as discussed in more detail in the section on Linking and Brushing The Exclude selected option is off by default and must be activated explicitly The other options Background Color Save Selected Obs and Save Image as work the same as for the other graphs and maps note how in Figure 112 the background color was changed to gray from the default white ScatterPlot gt N Standardized data Exclude selected Raw data Save Image as Save Selected Obs HI A Background Color E D a Figure 113 Scatter plot options AQ Linking and Brushing A distinguishing feature of GeoDa and its precursor DynESDA is the implementation of full two way dynamic linking This approach consists of treating each window statistical graph map table as a view of the data and connecting all
99. ox Plot by clicking outside the solid part in a Map A slightly more convoluted way to obtain the same result is to select all observations drag a rectangle around the whole Box Plot followed by Double Click to select the complement Selection in a Scatter Plot Selection in a Scatter Plot operates in much the same way as for the Box Plot Individual observations are selected by clicking on the matching points in the plot Groups of observations are selected by dragging a rectangle around them The selected points are highlighted in yellow default selection color Observations can be added to an existing selection by using Shift Click for individual points or by holding down the shift key while dragging a rectangle In general the selection status of any single observation or group of observations using a rectangle can be switched by means of the shift key 73 With the Exclude Selected option on any selection on a Scatter Plot results in the regression fit to be recalculated A new line is drawn in the graph in brown the original fit remains blue and the slope in the new line appears on the right hand side above the graph Figure 117 The new fit is computed for a data set without the selected observations To get the result for the selection only you need to Double Click to switch the selection to its complement The Exclude Selection option allows for the visual inspection of the effect of outliers and leverage points on a re
100. p gt Percentile A Percentile map is also a special case of a quantile map In this case no additional categories are created but the categories are grouped to accentuate the extreme values Specifically six legend categories are created corresponding to lt 1 1 to lt 10 10 to lt 50 50 to lt 90 90 to lt 99 and gt 99 This is illustrated for the 1979 SIDS death rates in North Carolina counties in Figure 63 This is invoked by the Map gt Percentile command without additional dialogs Percentile SIDR79 Percentile SIDR 9 lt 1 0 WY a 10 ao 10 50 40 mm 90 40 DZ am 29 9 gt 99 2 Figure 63 Percentile map for 1979 SIDS death rates in North Carolina counties Note some particular features in Figure 63 The parentheses next to the first category show no observations in this category while one would expect 1 NC has 100 counties In fact due to ties that first observation cannot be assigned to a separate percentile and is grouped with the 1 to lt 10 category Similarly at the high end there are two observations one for 99 and one for 100 due to the decision rule used to assign observations to categories Map Movie A Map Movie is an animated sequence highlighting locations on the map in increasing order of magnitude for a specified variable It may suggest spatial structure in the data in the form of spatial regimes For example a Map gt Map Movie gt Cumulative c
101. ravciscceassasdeanccaes stag a Aaa 42 EIERE eege Eed 42 Percentile map EE E NEE 43 Map color selection options ssesssseessesesseesseetsseessertsserssstesseressressseesseessseesseres 43 E ENER LCE 44 71 12 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 Customizing the selection hash marks color 44 Customizing the background color for a map View cccesceeceeseteeeeeeeteeeeeees 44 Save map as image ele ege 45 SUE BE 45 Save selected observations indicator variable dialog cccccceeseeeeeeeetteeeeeetees 46 Indicator variable added ER E 46 e ee Ee 47 Specifying spatial weights for rate smoothing eee ee eeececeeeeteeeeeeeteeeeeeeeaes 48 Raw rate box map for 1979 SIDS death rate in North Carolina counties 48 Excess risk map for 1979 SIDS death rates in North Carolina counties 49 EB smoothed box map for 1979 SIDS death rates in North Carolina counties 50 Spatial rate box map for 1979 SIDS death rates in North Carolina counties 50 Spatial EB rate box map for 1979 SIDS death rates in North Carolina counties 51 Save rates option for rate EEN 52 EE ENEE E EE 52 Computed rate variables added to data table cecceccceeeeseeeeeeeeneeeeeeeeeeeeeees 52 New rate variables available for analysis 0 cccccescccccsesteeeeeeeteeeeeese
102. re given in the order in which they appear in the shape file that is stored on disk which may not be the most useful Sorting of records according to the values taken by a given variable is done directly by double clicking on the field variable name at the top of the table as shown in Figure 89 for the POLYID variable in the Columbus data set The result is a table sorted by increasing values of 44 POLYID Note the small triangle pointing up next to the variable name Double clicking again toggles the order and sorts the records by decreasing value of the variable This is indicated by the triangle pointing down as in Figure 90 The sorted records can be restored to the original order by double clicking on the header for the left most column This column contains sequence numbers for the original order The sorting mechanism is applied to it in the same way as to any other variable hence the triangle appears and the order can be reversed double click to toggle the sort order Note that the sorted order of records is not permanent nor does it affect the way the data are stored on disk Table COLUMBUS AREA PERIMETER COLUMBUS_ COLUMBUS_I POLYID i 0 309441 2 440629 2 5 1 d 0 250329 2 236939 3 i 2 a 0 192468 2 187547 4 6 3 4 0 083841 1 427635 5 4 E 0 488888 2 997133 6 7 5 Figure 89 Records sorted by increasing values of POLYID Table COLUMBUS AREA PERIMETER COLUMBUS_ COLUMBUS_I POLY
103. rent shapes as illustrated in Figure 66 point rectangle polygon line and circle Selection by forming a Rectangle through clicking and dragging the mouse is the default The observations are selected following a spatial select rule which selects those locations whose centroid falls within the rectangle The other selection rules are self explanatory The selection rules can also be set by right clicking on a map A selection rule stays active until a different one is chosen Note that a selection rule is specific to each window so different windows different maps may have different selection rules such as Rectangle in one and Point in the other Options Window Help Selection Shape N d Point Zoom gt v Rectangle Color gt Polygon Line Add Centroids to Table Circle T Figure 66 Map selection rules Options gt Zoom The map view is not static but supports the usual zoom in and zoom out functions You invoke this as Options gt Zoom and select one of the three choices Figure 67 Alternatively you can right click on the map Zoom In requires you to draw a rectangle with the pointer to rescale the picture Figure 68 To Zoom Out you need to click on the map no rectangle Zoom gt Full Extent restores the original map view Options Window Help __ Selection Shape gt hl wil Poon o Color gt zoom Out Add Centroids to Table Full Extent Figure 67 Zoom options 4 m
104. rvations all 19 observations in quintile 1 and 2 would receive the color of quintile 2 Map gt Std Dev A standard deviational map groups observations according to where their values fall on a standardized range expressed as standard deviational units away from the mean The Map gt Std Dev command creates a choropleth map with the categories corresponding to multiples of standard deviational units No additional dialog is required For example a standard deviational map for the Columbus neighborhood crime data is illustrated in Figure 61 The map legend shows the type of map variable its mean value the break points for each category and in parentheses how many observations fall in that category As for the quantile map the legend colors can be customized for each individual category Std Deviation CRIME Std Deviation CRIME WA E mm 33 44 33 44 36 86 2 Mean 35 86 35 86 38 28 2 38 28 40 70 4 w gt 40 70 h Figure 61 Standard deviational map for Columbus neighborhood crime 4 A standardized variable has a mean of zero and a standard deviation of 1 by construction Hence a standardized value can be interpreted as multiples of standard deviational units 3R Outlier Maps Outlier maps highlight locations with extreme values both high as well as low GeoDa contains two types of outlier maps a Box Map and a Percentile Map These are choropleth maps and as suc
105. ry files as ascii with bounding box 114 Data Manipulation e Tools weights creation data conversion available without opening project e More flexible spatial weights construction for polygon shape files include distance for arbitrary x y coordinates centroid x y coordinates as default for distance computation e Thiessen polygon shape files include area and perimeter as variables
106. s for the complete data set In Figure 139 the background color has been changed to gray on Options gt Randomization Inference for Moran s I is based on a permutation approach in which a reference distribution is calculated for spatially random layouts with the same data values as observed The randomization uses an algorithm to generate spatially random simulated data sets outlined in Anselin 1986 It is invoked by Options gt Randomization gt xxx or by right clicking anywhere in the graph window The xxx is the number of random permutations used in constructing the reference distribution such as 99 or 999 as shown in Figure 140 pRandomization gt 99 Permutations nvelope Slopes ON 199 Permutations f 499 Permutations Save Results l 999 Permutations Save Image as Other Caria Calactad Nhe wm emgeet Figure 140 Randomization option in Moran scatter plot The result is a window depicting a histogram for the reference distribution with the observed Moran s I shown as a yellow bar as in Figure 141 for the Columbus variable CRIME In addition to the histogram a small number of summary statistics are listed as well In the top left corner the number of permutations used 999 and the pseudo significance level 0 001 are given The latter is computed as the ratio of the number of statistics for the randomly generated data sets that are equal to or exceed the observed statistic 1 over the number of permuta
107. s in Columbus 81 Construction of spatial weights based on rook contiguity eeeeeeeeeeeeeneeeeees 82 Completion of weights copnstructon 83 Specifying a higher order of contt gut 83 Distance weights creation EE 84 Specifying the threshold distance for distance band spatial weight 85 Specifying the order for k nearest neighbor spatial weights 0 0 0 0 eeeeseeeeeee 85 Selecting a weights file for the analysis of its characteristiCs cseeeeseeeeees 86 Connectivity of first order contiguity for North Carolina counties c 87 Rook connectivity for Iredell county NC ue ceetadseens 87 Weights selection dialog EEN 88 Moran scatter plot for Columbus CRIME cccceceeccccceseteeeeeeeeeeceeeeeeeeeesenneeees 89 Univariate Moran scatter plot option 90 Univariate Moran scatter plot with selected observations excluded 045 90 Randomization option in Moran scatter plot 91 Reference distribution for Moran s I for Columbus CRIME cceeceeeeeneeeeees 92 Reference distribution for Moran s I for Columbus open 92 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 Envelope slopes in the Moran scatter plot icons cacesssncacosnts Seven tass cendedsnscessencees 93 Save results dialog in Moran scatter plot 93 Moran scatter plot variables added to the data table ee eeeseeeeseeeesteeeeeeeeeees 94 Bivariate Moran scatter p
108. sidcent Save as type ESRI Shapefiles shp Ce l Figure 29 Shape conversion output save as file dialog SHAPE CONVERSION X Create Reset Done Figure 30 Shape conversion create button SHAPE CONVERSION DDPRREIEREIRDEIEDDDD EISE EISES EE OS EE Ceste Reset Rose ee Figure 31 Completed Polygon to Point conversion HI If you check the contents of the working directory you will notice three new files sidcent shp sidcent shx and sidcent dbf You can now add the point map to your project for example using Edit gt New Map or by clicking on the New Map button on the toolbar This is illustrated in Figure 32 for the North Carolina county sips data gt wap m sidcent ER Figure 32 Centroids for North Carolina counties The data table for the new point shape file contains all the variables from the original polygon shape file including AREA and PERIMETER which of course don t make sense for points You make the table visible by clicking on the Table toolbar button or using Explore gt Table from the menu The result is as shown in Figure 33 You can delete the two variables using the Table edit functions right click and Delete Column Note also how the table does not show an explicit field for the x and y coordinates for the centroids Since this is a point shape file the point coordinates are invisible just as the boundary coordinates are invisible in a polygon shap
109. st have a unique value for each observation It is used to implement the linking between the maps and the different statistical graphs GeoDa will try to guess a variable to be used as Key Accept the default or go down the drop down list to select the variable as in Figure 4 If the one you picked does not have a unique value for each observation an error message will appear If none of the variables in your data set takes a unique value for each observation you must add one to the data table before you can use GeoDa If necessary use a spreadsheet package or ArcView to add a unique variable to the dbf table In the Columbus example the Key Variable is POLYID click og to accept the default or alternatively click on the variable name followed by ok to select it or double click on the variable name After you select an appropriate Key Variable a map is produced with the outline of the shape file as in Figure 5 for the Columbus neighborhood polygon data At this point the complete set of menu items and toolbar buttons becomes available as well GeoDa Project Setting Input Map shp C Program Files GeoDa Sample COLUMBU Gam Key Variable Figure 4 Key variable GeoDa 0 9 Beta COLUMBUS File Edit View Tools Explore Map SE Window L alal olele ee el e m COLUMBUS rex Figure 5 Opening screen with selected shape file outline Menu Overview General Features The GeoDa menu bar contains nine menu ite
110. statistical principles and elementary spatial statistics A series of specialized tutorials on the use of GeoDa for exploratory data analysis spatial correlation analysis etc is available from the GeoDa web site http sal agecon uiuc edu csiss geoda html Further background on methodological issues can be found in for example Bailey and Gatrell 1995 or Lattice data are discrete spatial units that are not a sample from an underlying continuous surface geostatistical data or locations of events point patterns GeoDa currently does not yet contain specific techniques to analyze geostatistical or point pattern data Fotheringham et al 2000 as well as in the sources mentioned above An extended set of course notes and examples dealing with an introduction to spatial data analysis can also be found in the repository of CSISS materials and Learning Resources at http www csiss org learning _resources content syllabi gis GeoDa was first available as a prototype in October 2001 as DynESDA2 Its first public release was in February 2003 The version of GeoDa covered in this user s guide is 0 9 3 June 4 2003 A summary of changes from the first release Version 0 9 0 is given in Appendix D The software is available for free for non commercial use Please refer to the license agreement in Appendix A before installing the software The software is provided as is and does not come with support other than what is con
111. t hand panel the FIPS codes are used while in the right panel simple sequence numbers appear GAL files can be edited with any text editor The numeric values are separated by a single space which can be replaced by another separator through a global Find Replace command in a text editor RN di sidr GAL WordPad O X sidr4 GAL File Edit View Insert Format Help Deh 4a a an R SIDS FIPSNO Del 8k a 3 37193 37005 3 37171 37009 5 37197 37193 37169 37005 2 37029 4 37185 37091 37015 For Help press F1 For Help press F1 Figure 125 GAL format spatial weights file for North Carolina counties The cwrt format for spatial weights is used when the neighbor information is derived from the distance between points Again the result is a text file with a header line followed by the neighbor information When a Key Variable has been specified the header line is as in Figure 126 for k nearest neighbors of order 4 in the Columbus data set It contains four items 0 for future use the number of observations 49 the name of the shape file coLumBus and the Key Variable POLYID When no Key Variable is specified but sequence numbers are used the header line consists only of the number of observations The remainder of the file contains on each line the origin rp value i e the 1p for the observation the destination ID i e the 1p for the neighbor and the distance on which the contigu
112. t seng f icsccseseccesecsdascgaeuadisocesiescoseagsasaseebercatseaseanseebeseaes 13 Explore TMEV se eas syishis ekty a ia i ea E R a craven ava lav aad 14 HEH 15 Map Box Maps s bmeni nes ders eege Eege 16 Map Smooth submenu ee EE 16 Map Movie submentt c sincntstind atitandaies ction aii aviistinn diate anaes 17 Tee EE 17 Help Met EE 18 ERT EE 18 Shape conversion dialog ornoen nae See 20 Shape conversion input EE e E 20 Shape conversion input file thumbnail 0 0 ccc ececccecceesneeeeeeeneeeeeeeeeeeeeeeeaes 20 Shape conversion output save as file dialog c eececcccceseeeeeeeeteeeeeeseneeeeeseeaes 21 Shape conversion create button 21 Completed Polygon to Point conversion ccescccceeeeteeeeeeeeteeeeeeeeteeeeesenteeeeeees 21 Centroids for North Carolina coumnties 22 Data table for centroids point shape Die 22 Add centroids to table option in men 23 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 Add centroids to table from INIA EE 23 Add centroids to EE 24 Columns with centroids added to the sidcent data table eeeseeeeeneeeeeneeeeees 24 Export Centroids file dialog geet aac ne tease 23 Centroids dbf file loaded into spreadehect 25 Point to Polygon shape conversion dalog 26 Completed Point to Polygon shape cConveraion 21 Thiessen polygons for Baltimore point data with point data overla
113. tained in the input file Select the variable names for the coordinates and click the Create button The shape file will be saved to the working directory You can next open it in GeoDa As shown in Figure 46 the points and data table are identical to the ones for the Baltimore sample point shape file The procedure for Tools gt Shape gt Points from ASCII is similar Again you need to specify the input and output file name in a dialog Figure 47 as well as the variable names for the x and y coordinates If the input ascii file does not correspond to the comma delimited format or does not have the proper header line an error message appears you can use any text editor to make sure the header line is in the proper format Convert DBF to SHP Input file dbf C Program Files GeoDa S ample BALTIM DBF LS Output file D hei C Program Files GeoDa Sample baltimpointSHP RM coord STATION Se Y coord STATION md Figure 45 Points from DBF file and variable selection dialog 29 baltimpoint STATION PRICE NROOM DWELL NBATH PATIO 47 000000 4 000000 0 000000 1 000000 0 000000 1 2 113 000000 7 000000 1 000000 2 500000 1 000000 1 000000 2 165 000000 7 000000 1 000000 2 500000 1 000000 1 000000 4 104 300000 7 000000 1 000000 2 500000 1 000000 1 000000 In Figure 47 the input file is baltim csv and the output file is set as baltipoint2 shp Again x and y are the coordinates Click on the create butto
114. tained in the web pages and provided informally through the Openspace mailing list Some minor problems with the software can be expected Please report anything that seems like a bug to anselin uiuc edu or alternatively post to the Openspace mailing list mailto openspace sal agecon uiuc edu Note GeoDa replaces the DynESDA Extension for ArcView 3 x This extension is now deprecated and is no longer supported To subscribe to the mailing list check http sal agecon uiuc edu mailman listinfo openspace Getting Started Once you have installed GeoDa see Appendix C you can launch the program by double clicking on its icon or using any of the other standard MS Windows approaches The opening screen is as in Figure 1 You start the analysis of a data set by clicking on the Open Project button on the toolbar Figure 1 or by using the menu File gt Open Project Note that in order to use the data manipulation Tools it is not necessary to open a project since they operate on separate files Since the Project requires you to have data in a shape file you can use the Tools before the actual analysis to construct the required shape files for example to create a point shape file from X Y coordinates in an ascii file GeoDa 0 9 Beta File View Tools Help New Map Document Figure 1 GeoDa opening screen Project Setting The Project Setting dialog Figure 2 requires you to enter the file name of
115. teeeeeseaaes 53 Table MENU EE ENER 54 Records sorted by increasing values Of Sotetp 55 Records sorted by decreasing values Of Sorten 55 Editing individual table ele Geesse eege ege ee 56 Manual record selection in E 56 Selected records promoted to the top of the table cc eececceeesseceeeeneeeeeeaes 56 Range selection interval specification c cccccccssseceesssteceeeeeeeeeeeeseseeeeenenaeeees 57 Range selection applied to table kee dE 58 Specifying a regime variable following a range selection ss ssseseessseressseeee 58 Regime variable Bee 58 Save selected observations in Table 2oek7ueedeehegNEERZeeEN NEEN ENENe deen 59 Selected observations indicator variable added to data oble eee 59 Specifying a variable name for a new column field 000 0 eeeeeeeeetteeeeeeeeee 60 Field calculation unary operattong 60 Field calculation binary operations cccccccccessececeeeeseeeeeeeeeeceeseeneeeeeseneeees 61 Computing a spatially lagged variable cceecccccceesseeeeeeeeneeeeeeeeneeeeeneeeeeeeees 62 Rate Calculation in a table aiscvaisvcinsntavsinasdasaceonteanrasnbauaieavtacdinardaay eenig 62 Joi tables dialo b cede nie See E 63 Joined variables added to data table ceesceesseeesseeeeeneeeesneeeeeseeeeesneeeseneeeeees 64 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136
116. the Approximate Nearest Neighbor Library ANN Permission to use copy and distribute this software and its documentation is hereby granted free of charge provided that 1 it is not a component of a commercial product and 2 this notice appears in all copies of the software and related documentation ANN may be distributed by any means provided that the original files remain intact and no charge is made other than for reasonable distribution costs Disclaimer The University of Maryland and the authors make no representations about the suitability or fitness of this software for any purpose It is provided as is without express or implied warranty Version 0 1 03 04 98 Preliminary release Version 0 2 06 24 98 Changes for SGI compilation Authors David Mount Dept of Computer Science University of Maryland College Park MD 20742 USA mount cs umd edu Sunil Arya Department of Computer Science The Hong Kong University of Science and Technology Clear Water Bay Kowloon Hong Kong arya cs ust hk Appendix C Installing GeoDa GeoDa comes with a simple installation script Locate the setup exe file and double click or use Run from the start button and type the path to Setup exe The Installshield script will open a welcome window Type in the required information and the program will copy all the necessary files A typical installation will move GeoDa exe to the c Program Files GeoDa directory There are
117. the text box of the Save As dialog followed by clicking on the Save button as shown in Figure 123 Once the two files are specified their path names will be listed in the text box of the creating Weights dialog In addition all the options in the creating Weights dialog become active that is they change from being grayed out to being visible 7R CREATING WEIGHTS po Reen Figure 122 Creating weights dialog CREATING WEIGHTS C Program Files GeoDa Sample COLUMBU ta lt Rec_Num 1 2 3 N gt e ri Save As P colgali GAL Figure 123 Wegen file giton for PET ach In addition to the input and output files a Key Variable must be specified that corresponds to the ID values used in the weights files Figure 124 This can be the same Key Variable as used in the project but this is not required as long as the variable takes on a unique value for each observation If no Key Variable is specified the default is 79 the sequence number of the observation Using this default is usually not a good idea since the order of the observations may be different between different file formats e g shape file vs ascii output file CREATING WEIGHTS Ed Input File shp IC Program Files GeoDa Sample COLUMBU G Save output as C Program Files GeoD a ample colrook gal Select an ID variable for the weights file lt Rec_Num 133 Ms E lt Rec_Num 13223 MS a COLUMB
118. these views together The connection is such that any time an observation is selected in any view it is simultaneously selected highlighted in all the other views This allows the user extensive interaction with the data and is a fundamental tool in EDA GeoDa implements the selection of individual observations or groups of observations for all statistical graphs and maps as well as for the table Any time more than one view is open selection automatically leads to linking The linking is dynamic in the sense that any change in the selection in any of the views directly leads to an update of the selection in all the other views Brushing is a slightly more involved form of dynamic linking in that a brush is moved over the observations in one of the views In GeoDa the brush is a rectangle specified by the user As the brush moves over the data the selection changes and is updated in all the other views in the project When the Exclude selected option is set in the Scatter Plot brushing results in an instantaneous recalculation of statistics in a Scatter Plot such as the slope of the regression line and Moran s I spatial autocorrelation statistic In each of the maps and graphs an indicator variable can be created and added to the Table using the Save Selected Obs function in the Options use the options menu or right click on the graph or map This indicator variable will take on the value of one for the selected observations and
119. three statistical graphs The graphs are dynamically linked allowing for full brushing of the data see the section on Linking and Brushing They are invoked from the Explore menu see Figure 18 or by clicking on the corresponding Explore toolbar button The three statistical graphs included in GeoDa are the Histogram Scatter Plot and Box Plot They require that a shape file is loaded into the project and that a variable be specified using a Variables Settings dialog see Figure 12 The Histogram and Box Plot require one variable to be selected the Scatter Plot two If these variables were set earlier as default there will be no Variable Settings dialog when the statistical graphs are invoked To change the default variable it is necessary to explicitly reselect it using Edit gt Select Variable or by clicking the Select Variable toolbar button Explore gt Histogram Creates a Histogram for the specified variable The default number of classifications is set to 7 This can be changed by invoking Options gt Intervals which will bring up a dialog where a new value can be specified for example 12 in Figure 107 Clicking on the ox button will create a new Histogram with the specified number of intervals For example using the Columbus neighborhood shape file coLUmBUS SHP a histogram of housing values HOVAL is given in Figure 108 The category break points are listed on the right hand side and the number of observations in each of t
120. tics In GeoDa they are also used to implement Spatial Rate smoothing Weights can be constructed based on contiguity from polygon boundary files the original layout or a set of Thiessen polygons or calculated from the distance between points points in a point shape file or any x Y coordinates in a file Opening Existing Weights Existing spatial weights are added to a project by invoking Tools gt Weights gt Open or by clicking on the Open weights matrix toolbar button This creates a Select Weight dialog as shown in Figure 119 When no weights have been opened for the current project the second radio button is checked by default This forces you to select an existing weights file in either GAL or Gwr format see next section Click on the open File icon to obtain the usual dialog with the contents of the current working directory your contents will likely differ from those shown in Figure 119 Select the weights file and click on Open to add it to the Select Weight dialog Finally click ox to add it to the project as in Figure 120 Once a weights file has been opened it becomes available for any spatial analysis Invoking the Tools gt Weights gt Open at this point brings up the Select Weight dialog with the first radio button checked and the available weights given in a drop down list as shown in Figure 121 At this point you can also still check the other radio button to open a new weights file SELECT WEIGHT X
121. tions used 1 Hence the value of 0 001 in Figure 141 indicates that none of the simulated values were larger than the observed 0 52 In the bottom left of the graph the value for Moran s T is listed its theoretical mean E I the average of the reference distributions Mean and the standard deviation for the reference distribution Sd You can generate another set of simulated values by clicking on the Run button Clicking Close removes the Randomization window An example of a reference distribution where the statistic does not turn out to be significant is shown in Figure 142 for the Columbus variable open Moran s I is 0 067 with pseudo significance of 0 31 Note how the mean computed for the reference distribution 0 0181 is slightly different from its theoretical value 0 0208 91 Randomization Figure 141 Reference distribution for Moran s I for Columbus CRIME Randomization Figure 142 Reference distribution for Moran s I for Columbus oPEN Options gt Envelope Slopes ON A different way to visualize the range of autocorrelation statistics that can be obtained in spatially random simulated data sets is shown with Options gt Envelope Slopes ON As for the other options it is invoked by using the Options menu or by right clicking anywhere in the scatter plot window This plots the lower 5 percentile and upper 95 percentile of the reference distribution as the slope of dashed lines in the Moran Scatter
122. two additional directories C Program Files GeoDa Samples which contains the sample data sets and C Program Files GeoDa Docs which contains this document A shortcut will be created as well and appear as an icon on your desktop At the end of the installation you can check the box to start the program before clicking on the Finish button Installation Issues There have been some problems with incompatibilities between various MS Windows systems libraries on Win98 and Win2000 platforms Before reporting these problems make sure the files listed below are present in two important directories If some files are missing they may have been copied to the wrong directory by the install program Try to locate them on your system and if found copy them manually into the appropriate directory this is sometimes the case with misplaced ESRI MapObjects LT libraries Required Files in the Directory c Program Files Common Files ESRI files in bold must be registered on the target computer the installation program should take care of this but you can always do it manually as well refer to the Windows Registry instructions for your system 04 05 2000 01 15 PM 520 192 AFLT20 dl11l 04 05 2000 01 21 PM 380 928 MOLT20 ocx 04 05 2000 12 51 PM 622 592 Pe dl 04 05 2000 12 51 PM 248 320 Sg dl 04 05 2000 01 20 PM 77 824 ShapeLT20 d11 Required System Files The following files should be in the systems directory for your flavor of the Windows
123. ual location will select all other locations i e the full map except that location This may have unexpected effects on individual graphs when the Exclude Selected option is on in the sense that a regression will be computed for a single observation Selection in a Histogram In a Histogran it is not possible to select individual observations only groups of observations contained in the same category A category is selected by clicking on its bar or by clicking on the small square shown next to the breakpoints on the right hand side of the window Additional categories may be selected by using Shift Click In general the selection status of a category can be turned on or off using Shift Click Selected 71 categories do not have to be adjoining in the Histogram A Double Click anywhere in the Histogram will reverse the selection to its complement the selected will be unselected and vice versa The selected category or categories will be highlighted in yellow as shown in Figure 115 for the CRIME variable in the Columbus neighborhood data set COLUMBUS SHP m Histogram CRIME DEAR E selected features Oo 0 178269 5 9044169 fies 5 90441 69 11 650565 E 11 6 0565 17 556713 E 17 956715 25 082061 LJ 23 082861 28 609009 26 609009 54 535156 im 4535156 40 261 04 El KEE EH Sj 45 987452 51 71 6 DEI E 57 4 9744 63 165896 65 165696266 692004 CRIME Figure 115 Histogram with three highest categories selected
124. uantile maps This map is illustrated in Figure 81 for a Box Map of the 1979 SIDS death rates Note the contrast with Figure 79 there is now one outlier county at the low end of the scale and five outliers at the high end 49 BoxMap Hinge 1 5 EBS Smoothed SID79 w BIR79 Boxhtap Hinge 1 5 EBS Smootl Lower outlier 1 lt 25 25 25 50 25 50 75 25 gt 75 19 Upper outlier 5 Figure 81 EB smoothed box map for 1979 SIDS death rates in North Carolina counties Map gt Smooth gt Spatial Rate Creates a smoothed map using a spatial window average The smoothed rate is computed from the ratio of the total number of events in a spatial window to the total population at risk in that window The window is specified using a spatial weights file and includes both the neighbors as well as the location itself This is illustrated in Figure 82 with a Box Map for the 1979 SIDS data using a first order contiguity rook spatial weights file to define the window The Spatial Rate smoother brings out broad spatial trends in the data and typically greatly reduces the number of outliers The map header lists the type of smoothing and the variables used in the calculation m BoxMap Hinge 1 5 SRS Smoothed SID79 w BIR79 Boxhtap Hinge 1 5 SRS Srmoot Lower outier 1 E 25 25 25 50 25 om 75 25 E zen 24 Upper outlier 0 Figure 82 Spatial rate box map for 1979 SIDS death rates
125. ult for the selection in Figure 92 would be as shown in Figure 93 m Table COLUMBUS 2 440629 2 236939 1 427635 5 2 997133 6 7 Figure 92 Manual record selection in a table Table COLUMBUS 3 BPE 4 3 i 2 440629 2 1 al 2 236939 3 2 4 1 427635 5 4 el 2 997133 6 5 Figure 93 Selected records promoted to the top of the table 54 Clear Selection Any selection can be undone by choosing Clear Selection for the table menu The removal of the selection affects all statistical graphs and maps in the project see the section on Linking and Brushing Range Selection The table contains some limited functionality to carry out queries where records are selected as the result of a logical operation comparing the value of a variable field to a numerical range This is invoked by Range Selection in the table menu which brings up a dialog to specify the parameters for the range interval as in Figure 94 The drop down list and set of text boxes under the heading Range Selection pertains to the interval itself Note how in the current version the range is inclusive of the lower bound lt and exclusive of the upper bound lt For example using the Columbus sample data set the records with a CRIME rate over 30 can be selected by specifying 30 as the lower bound and a large number larger than any in the data set as the upper bound as in Figure 94 In order to implement an equality co
126. ven weights file RA Connectivity of sidrook GAL sips EX Eeee Ci et Lk Ei m E E E Connectivity Figure 134 Connectivity of first order contiguity for North Carolina counties Connectivity of sidrook GAL A SIDS Table SIDS OX AREA PERIMETER CNTY_ID STATE_NAME STATE_FIPS CNTY_FIPS il D 1947 1947 Iredell North Carolina 37 097 4 0 114000 1 442000 1825 1825 Ashe North Carolina 37 Jeff E Kiew 4 2 1090 1090 EN wru larha Caralina 27 4174 a Figure 135 Rook connectivity for Iredell county NC R7 Global Spatial Autocorrelation Global spatial autocorrelation analysis is handled in GeoDa by means of Moran s I spatial autocorrelation statistic and its visualization in the form of a Moran Scatter Plot Anselin 1995 1996 The Moran Scatter Plot is a special case of a Scatter Plot and as such has the same basic options It is linked to all the graphs and maps in the project allowing full brushing Global spatial autocorrelation analysis requires that both a variable be specified using the variables Settings dialog as well as a spatial weights file The latter is specified through a dialog that asks to select either a new file from disk or one from a list of currently in use spatial weights as the result of an earlier Tools gt Weights gt Open command as in Figure 136 SELECT WEIGHT e Select from currently used C Program Files Geo
127. voked by Explore gt Multivariate LISA or by clicking on the Multivariate LISA button in the Explore toolbar The same four graphs can be generated as for the Univariate LISA except that they pertain to a bivariate measure of local spatial autocorrelation All options are the same as for the Univariate LISA EB LISA The Graes principle can also be applied to an EB standardized rate variable This operates the same as for the standard univariate measure of local spatial autocorrelation except that the variable specification dialog asks for both Event and Base variables as in Figure 149 It is invoked by Explore gt LISA with EB Rate or by clicking on the matching button on the Explore toolbar The same four graphs can be generated as for the Univariate LISA except that they pertain to a measure of local spatial autocorrelation computed for EB rates All options are the same as for the Univariate LISA 104 References Anselin L 1986 MicroQAP a Microcomputer Implementation of Generalized Measures of Spatial Association Working Paper Department of Geography University of California Santa Barbara Anselin L 1994 Exploratory Spatial Data Analysis and Geographic Information Systems In M Painho ed New Tools for Spatial Analysis Eurostat Luxembourg 1994 pp 45 54 Anselin L 1995 Local Indicators of Spatial Association LISA Geographical Analysis 27 93 115 Anselin L 1996 The Moran Scatterplot as an
128. yed 27 Data table for Thiessen polygon shape Die 28 Baltimore point data as a comma delimited csv Die 29 Points from dbf file and variable selection dalog 29 Baltimpoint point shape file created from dbf input 2 0 0 0 eee eeeeeeeeeeeeeeeneeeeees 30 Points from ascii file and variable selection dualog 30 Baltipoint2 point shape file created from ascii mput ce eeeeeeeeeeesteeeeeneeeeees 31 Boundary file format left and format la Geht 32 Boundary file format 2 left and format 2a Geht 32 Bounding box coordinates for Columbus polygons cccesceceeeeeteeeeeeeeeeees 32 Polygon shape to boundary file export dialog eee eeeeseeeesteeeeeeeeseneeeesneeeees 33 Ascii text output file from Columbus data set polygon shape file 33 Exporting data dialogs sesioen nis raine EEE a aE 34 Data teady for EE 34 Exposing the Resend E 35 Quantile Map At alO Osc ated aa as aed Me ee hed tees ceeds Tet Aes 36 Columbus neighborhood crime quintile map 36 Legend color customization reit Eeer Eege 37 Customizing JeeentzGerhte stage eegend geed Tee EES EE 37 Standard deviational map for Columbus neighborhood crime ssssesseseesseeessee 38 Box Map for 1979 SIDS death rate in North Carolina counties ceeeeeee 39 Percentile map for 1979 SIDS death rates in North Carolina counties 40 KEE e EE EE 4 Map options and shortcuts context menu 4 Map sele tion rules cise scavcaseciavciceciavcvate
129. ygon shape file SE Oo On Ek w N rz Creating Point Shape Files from DBF and ASCII Input Files While GeoDa requires project input files to be in the shape file format the Tools functions contain means to construct point shape files from input files that contain the x and Y coordinates as fields Tools gt Shape gt Points from DBF takes a dBase file as input while Tools gt Shape gt Points from ASCII reads a comma delimited ascii file The latter follows a specific format starting with a header line with the number of observations and the number of variables The second line contains the variable names separated by a comma Next follow the actual data with each row corresponding to an observation and the values separated by a comma Figure 44 illustrates the layout of the comma delimited baltim csv file corresponding to the BALTIM shp sample point data set This file was constructed by exporting the dbf file as csv in the Excel spreadsheet and adding the header line in any text editor Note how the file in Figure 44 contains the x and Y coordinates as variables DR E baltim csy Notepad File Edit Format View Help 211 17 STATION PRICE NROOM DWELL NBATH PATIO FIREPL AC BMENT NSTOR GAR AGE CITCOU LOTSZ SQFT X Y 1 47 000000 4 000000 0 000000 1 000000 0 000000 0 000000 0 QO0000 2 000000 3 C00000 0 000000 2 113 000000 7 000000 1 000000 2 500000 1 000000 1 000000 1 O00000 2 OD00
130. zero for the others Several such indicator variables may be constructed using different selection criteria Selection in a Map Selection of an individual observation or a group of observations is implemented slightly differently for each type of graph In a Map selection is based on the option set by the Options gt Selection shape command Figure 66 or by using the pop up menu right click in the map window itself You select a location by clicking on it Point selection by clicking and dragging to create the desired shape Rectangle or Circle or by consecutive clicking to form a Line or Polygon For the latter two the selection must be finalized by a Double click The other selection mechanisms do not require this 70 The selected locations following a spatial select rule are highlighted using a cross hatched pattern see Figure 114 The default color of the selection is yellow but this can be changed using the options command for example as in Figure 71 A click outside the solid part of the map clears the selections A Double Click outside the solid part of the map selects all locations COLUMBUS Figure 114 Selected locations on the Columbus neighborhood map using Line Select For a Point selection additional items may be added to the selection using shift Click More generally Shift Click will change the selection status of individual locations turning them on or off A Double click on an individ

GeoDa™ 0.9 User's Guide - The University of North Carolina at

Contents

Download Pdf Manuals

Related Search

Related Contents

GeoDa&trade; 0.9 User's Guide - The University of North Carolina at

Contents

Download Pdf Manuals

Related Search

Related Contents

GeoDa™ 0.9 User's Guide - The University of North Carolina at