Guide To Using the Crime Analytics for Space-Time
Contents
1. 6 6 Specialized Maps and Graphs 19 Jo Cluster Maps s ee pne y emn PM 9 6 2 Calendar Maps Los i R09 ala Ube bea 21 Trend ano eB she yD ene He qe Gr Ge ant ada nae ee 21 Outlook 27 List of Figures 1 CAST on Windows Mac OSX and Linux Ubuntu Main CAST Memili se bocoi bes a ee emn Static and Dynamic Map Menu Options pg P 4 Map Legend Dropdown Options 0 0 0 0 000000000008 Map Legend Color Dropdown Option Changing the Order of Map B Multiple Layers of Heatmaps for Four Crime 1 linked Views eee AA 11 Linking and Brushing sao capser n RR dh RE RR 2 Static and Dynamic i3 USA Cluster 220 sace eo ee GS eo ye 14 Static and Dynamic Calendar 15 tatic and Dynamic Trend Graphs 6 Trend Graph to Identify Periods with Significant Concentrations of Crime 26 1 Purpose The purpose of this guide is to2. coordinates in ArcGIS 10 project the shapefile and add two new fields such as pr and Y pr to the table float e g 20 10 for precision and scale Then right click on each of these fields and Calculate Geometry based on X and Y respectively for the desired projection Next import the dbf file back into ArcGIS through File Add Data Add XY Data and select the X pr and Y pr fields Save as a new shapefile at which point you can load both the point and area shapefile with the same projection in one view 3 2 Date Fields CAST allows analysts to aggregate point data on the fly to areas and to specified time periods and to save the space and time aggregated data as a new dbf To analyze crime over time a date field needs to be specified For point shapefiles each record will have a time stamp often with the date and time for the San Francisco sample data use the Date and Time fields You can specify custom formats for these date and time fields in CAST but for automatic detection of the date field it is easiest to have a separate field for the date and time of an event For instance if the data field is recorded as 05 01 2011 00 00 00 then analysts should turn this into two separate fields such as date 05 01 2011 and time 00 00 00 before loading the data into For polygon shapefiles without point data crimes need to already be aggregated for specific time periods one period per column 3 3 Data Used
3. d Mates pow amet EX pe Wess vm Br sms sexi um onse MT Eel noo crime2012 projData July Dec2012 drugs x a hoang Mes Heime2012_projoars LI a Linked Views Across Scales oo ar ri b Selected Areas in Map Boxplot and Histogram Figure 10 Linked Views 17 i 3r E 4 al b Maps and Scatterplots after Selection Figure 11 Linking and Brushing 18 local Moran s I and 3 local G statistics maps for area data based on Getis Ord Figure 12a shows the settings and results for a heatmap where all crime incidents are pooled across time The settings for the dynamic heatmap are displayed in Figure 125 The interface for dynamic cluster maps contains a series of micromaps for each specified time period such as 1 month intervals below a larger map of the respective time period that is highlighted in the micromaps Figure Eg on the main map Further the animation can be started paused and stopped through the The micromaps can be displayed or hidden by right clicking controls in the toolbar see Figure 3 for all dynamic maps and graphs including calendar maps and trend graphs For the dynamic cluster maps analysts can also manu
4. robbery and vandalism while the association between robbery and car theft is much weaker The next Figure shows how the regression slopes and statistics are recomputed the fly when a subarea in the map is selected The red slope and numbers correspond to selected observations the blue ones to unselected ones and the pink to the area as a whole This suggests that the relationship between robbery and vandalism is stronger in the selected than unselected areas while the reverse is true for robbery and drugs The Chow test below the statistics at the top indicates whether the difference between the selected and unselected slopes is statistically significant the p value of 0 000 in all three cases suggests these differences are indeed larger than expected Brushing using the brush icon allows the dynamic selection of observations 5 4 Dynamic Maps and Graphs To visualize crime patterns in space and time CAST offers dynamic views of cluster maps calendar maps and trend graphs Three types of cluster maps are implemented in CAST 1 heatmaps for events based on kernel density estimation LISA maps for area data based on 16 eon CAST alpha NE amp 900 robbery drags vandal drugs eese Wap crime2032 orejDua July Dec2012 vandalism ex Bion x Bi
5. twitter feeds which are linked to its Facebook page Users of our NIJ funded software can also utilize the Center s Openspace listserv to address technical questions about these programs 15 At https geodacenter asu edu software 16 listserv https geodacenter asu edu support community Twitter http twitter com GeoDaCenter and Facebook http www facebook com geodacenter 28
6. Guide To Using the Crime Analytics for Space Time CAST Desktop Software Program Sergio J Rey Luc Anselin Xun Li and Julia Koschinsky February 9 2013 This program was supported by Award No 2009 SQ B9 K101 awarded by the National Institute of Justice Office of Justice Programs 0 5 Department of Justice The opinions findings and conclusions or recommendations expressed in this publication are those of the author s and do not necessarily reflect those of the Department of Justice questions or comments about this guide lia koschinsky please contact Julia Koschinsky at ju Contents 1 Purpose 4 2 verview T 4 6 d en uc ku at O aes ce o cuia e BERG disc a a airs 6 3 2 Data Fields s sy s VE Sou mr guide ede dress 7 3 Data Used in this Guide 2 2 2222 o ES 7 4 Main Menu and Context Options LI Toolbars so mh bee rhe ene gg 7 4 2 ontext Specific Dropdown Menus 0 4 3 ery Window e cesa poe g Eck A dens eve 1 5 Functionality 12 5 Multiple Layers SR RC ben So ml d 2 2 Separate Views with Different Shapefle 2 5 3 Linkimg and Brushing soss s 29 e oo o o o ok o ok oh n ko n n RR RE 5 4 Dynamic Maps and Graphs
7. ally move the micromaps forward or backward by clicking to the left or right of a highlighted micromap in all map views additional layers can be added that will be shown in the large map view 6 Specialized Maps and Graphs 6 1 Cluster Maps CAST contains two types of cluster maps for area data that are based on statistical tests local Moran s I and local G statistics The local Moran s I test that the Local Indicators of Spatial Association LISA maps are based on assesses to what extent values in one area are similar to the average value of neighboring areas CAST applies this test to compare he observed value of Moran s I to a spatially random reference distribution of Moran s I values The LISA map displays four significant options Hotspots high values in a given area surrounded by high values in a neighboring area coldspots low area values near ow neighboring values and two so called spatial outliers low values surrounded by high and vice versa The local G statistics maps display significant hotspots and coldspots but no spatial outliers There are two options for these maps Gi and Gi The Gi map identifies neighboring areas with average values that are significantly higher hotspots or ower coldspots compared to all areas The Gi map finds areas and its neighbors with 19 DE configuration Kernel Density Cel size Bandwidth eoo Map 90106 083107 v OR o
8. am see the W icons When a spatial weights file is created a text file is saved that lists the IDs of all the observations that are considered neighbors of a given observation The criteria for defining neighbors in CAST include contiguity shared borders rook and corners queen and distance within a specified radius or the closest fixed number of neighbors Details about weights creation and the weights histogram can be found in chapters 15 and 16 of OpenGeoDa s free tutorial 10 There are three types of map options to detect concentrations of crimes 1 general maps 2 calendar maps and 3 cluster maps General maps include options to identify unusual areas e g box maps in analogy to boxplots percentile maps that highlight the bottom and top one percentiles of a sorted distribution of crimes and standard deviation maps as well as typical map classifications based on natural breaks unique values and equal intervals The Map menu or icon option also allows analysts to create rates and rate smoothed maps to adjust for small underlying population bases that distort risk estimates Details abou rates and rate smoothing can be found in chapters 15 and 16 of OpenGeoDa s free tutorial and in PySAL s online documentation The calendar map and discussed below in more detail cluster map op Three trend graph options are available 1 A regular trend graph for user speci periods and areas 2 a dynamic tren
9. d graph which links a regu 10The OpenGeoDa tutorial can be downloaded at https geodacenter asu ed 1lFor more technical detail about the map classifications implemented in documentation at http pysal geodacenter org 1 5 library esda mapclassify The OpenGeoDa tutorial can be downloaded at https geodacenter asu ed PySAL s online documentation on rate smoothing http pysal geodacenter org 1 5 users tutorials smoothing html 9 ar trend graph t u system files geod html ions are fied time a map aworkbook pdf CAST see the online PySAL u system files geod aworkbook pdf is available at Map Legend Dropdown Options Change the map s transparency to view multiple layers Open the associated dbf table and view linked observations Zoomto the full extent of the map Cluster Map Local Moran s I See http pysal geodacenter org 1 4 users tutorials autocorrelation html i 424 For map classification details see http pysal geodacenter org 1 4 library esda mapclassify html Removes the map classification Figure 4 Map Legend Dropdown Options and then automatically displays crime patterns across time and space and 3 a significant trend graph which identifies statistically significant time runs within an area compared to the overall trend in this area The final set of tools include standard non spatial descriptive statistical views of the data namel
10. ence a selection in both map views is associated with Observations within the selection rectangle are selected The selection of areas is based on the centroid of an area 15 a selection of the same records in their shared table However the other shapefiles grids and two point files are each associated with their own tables The respective observations inside the selected spatial extent are highlighted in each of the three tables To compare which observations from the different shapefiles fall within the same selected subset the four tables can be opened and displayed next to each other In Figure 10 the distribution of the four crime types is visualized with the help of the four boxplots that are linked to a map and histogram although the boxplot scales are not Uu andardized The six districts with the highest frequency of robberies are highlighted in he boxplot to see how they compare to the other three crime types They coincide with the top ten areas for drug offenses and overlap with higher incidents of vandalism but the areas with the highest car thefts are generally elsewhere Figure 11a shows percentile maps of four offense types in San Francisco Robbery vandalism drugs and vehicle theft The catterplots below assess the strength of the statistical relationship between robbery and he other three offense types Robberies and drug offenses are most strongly correlated followed by
11. in this Guide data used in this guide represent crime incidents obtained from San Francisco Police De partment s SFPD Crime Incident Reporting system through San Francisco s open data por talf It includes four separate shapefiles of incidents of robberies sf robbery drugs narcotics possession or sale sf_drugs vehicle theft sf cartheft and vandalism sf vandalism for July 1to December 31 2012 and one polygon shapefile for SFPD reporting plots that these data were aggregated to sfpd plots The date and time fields to use are Date and Time in CAST 8 ArcGIS the date part in this example can be extracted through Left Date 10 or through DatePart Otherwise any program to parse text will work in this example the space indicates the break line 9The formatted data ready for CAST and metadata can be downloaded at http geodacenter org downloads data files SFCrime July Dec2012 zip N CAST alpha ioj x File Tools Table Heat Maps Events Local Moran Cluster Map Areas Gi Cluster Map Areas General Maps Boxplot Histogram Scatterplot Open Shapefile Open Movie Close All Views Open Table Create Spatial Weights for Cluster Maps View Weights Histogram Trend Graph Dynamic Trend Graph Significant Trend Graph Calendar Map Dynamic Calendar Map Figure 2 Main CAST Menu 4 Main Menu and Context Options 4 1 Menus and To
12. int outline color and to hide or display these outlines for improved pattern detection The order of layers can be changed by dragging and dropping one layer on top of another Once this occurs the dropdown menu shown in Figure 6 becomes available which grants the choice to move the dragged layer above or below the existing one 4 3 Query Windows Figure 7 presents the query window where analysts can specify the date and optional time fields for their point data and based on this define the date start end and time intervals for aggregating the data Filters for time of day and crime type also exist The polygon 11 m Figure 6 Changing the Order of Map Layers file that is selected in the Space field is used to aggregate points to areas The dbf file with the aggregated points by time and space can be saved and used elsewhere e g to access OpenGeoDa s time enabled features For cluster maps the query window also contains the option to specify a spatial weights file 5 Functionality 5 1 Multiple Layers As is common in Geographic Information Systems CAST allows analysts to load multiple layers in one view Figure B shows examples of how multiple layers can be used in CAST for instance to display incidents of robberies with percentile maps heatmaps and cluster maps 5 2 Separate Views with Different Shapefiles One of the special features in CAST is that analysts can
13. load multiple shapefiles at the same time in separate views that each contain several layers e g separate point shapefiles for robberies burglaries and vehicle theft each with area shapefiles as second layers or density 12 2 If a time field is selected events specified time frame can be chosen in military time the default is set to midnight 11 59pm 1 Select a date field and optional a time field Time of Day 00 00 23 59 8 3 In the Step By field you can identify the time steps you want to aggregate the events for Step By Month 6 4 The time interval defines the start and end date of your query By default the first and last Interval Start 07 01 2012 days in your data are displayed 5 Select one of the area shapefiles that you opened in CAST before from the dropdown or open a new area shapefile 6 If you want to filter your data specify the field and subcategory to do so optional 7 The weights file identifies neighboring areas If you do not already have a weight file create USA setting new one W icon Weights Create menu Weights Users julia sfpd_plots gal 7 contains the fields from this area shapefile plus the events aggregated to the time periods you specified and then select it here with this new dbf instead of the old one with OpenGeoDa s time enabled views 4 8 After selecting Run you have the option to sa
14. local Moran s I test will compare the crime count in a given time period e g the week of Dec 25 31 2012 to the average count of the previous 5 weeks to determine if this relationship differs from the average of all time periods The selection of neighbors in time creates a so called time weights matrix in analogy to he spatial weights matrix This time weights matrix identifies as many previous or past time 23 um September Uctober a Static Calendar Map b Dynamic Calendar Map Figure 14 Static and Dynamic Calendar Maps 24 Tend Graph robbe drugi vandal i lolol ssl ial eec rand Grate fd cathe robbed vandal a Static Trend Graph penny Eom Besson Trend Graph pots crhef robbery drugs vandal b Dynamic Trend Graph Figure 15 Static and Dynamic Trend Graphs 25 Ped i Time Pad 2 Time of Day 0000 2359 MOO Time steps to aggregate events for at least 20 periods from start to end date Specify of past and or future periods Select area s considered as neighbors for test a Steps Before Using a Significant Trend Graph c High High Time Run with One 5 Week Period Highlighted Trend Graph to Identify Periods with Significant Concentrations of Crime 26 periods as neighb
15. olbars Figure 2 shows the floating menu bar that is displayed when CAST is first opened In contrast to a typical Geographic Information System this menu bar is not associated with a fixed window to allow analysts to more flexibly open as many views as needed possibly across multiple screens without the constraint of a bounding box The starting point is the open folder icon or File Open Shape File menu option to open a shapefile Other file options include open map movie if you previously saved an animated gif file in CAST and close file to close all open views The Tools Menu also includes the option to create a grid with the same spatial extent as another shapefile This enables analysts who have crime events but no area shapefile to create a polygon shapefile that can be displayed in the same view as the crime incidents The table option can be chosen to view the dbf table that is associated with any open shapefile This is useful for viewing attribute details of incidents or areas selected in a view Next are tools to create so called spatial weights and view the number of neighbors of x Gl Hx alk pkk Time Slider Select 1 7 2012 31 12 2012 current 1 1 Month period Start and End Date Layers Add image pause Current marks position of slider Remove Stop Full Extent Open Table Brush dynamic selection Figure 3 Static and Dynamic Map Menu Options an area in a histogr
16. ors as specified that immediately precede or succeed a given time period For instance using the highlighted line in Figure 165 an example the crime counts in each week represent nodes that are connected through a trend line The time matrix identifies the five previous nodes for each given time period The Moran s I test is based on this neighbor definition to compare the count in the given period to the average count in the previous five periods As in the color coding for LISA maps those time runs that do differ from the average are shown as high high above average in a given time period compared to the previous and or future ones low low same for below average low high below average in a given period compared to above average for other periods and high low vice versa As in the spatial case the Gi and Gi tests compare the neighbors of a given period to the overall average or the neighbors and the given period to the overall average respectively For instance in the example below the Gi test assesses if the crime counts in the five previous time periods are above or below average compared to that of all time periods in an area The Gi test also includes the crime counts in a given period in addition to the previous five periods In the time case Gi and Gi also only identify hot and cold time runs not spatial outliers high low or low high as in the LISA case described above The local cluster maps only vis
17. previous software development efforts in cluding GeoDa and STAR PySAL framework provides a flexible background software engine for delivering a range of functionality customized to specific applications and using different user interfaces These applications include free standing desktop software such as 1 can be downloaded at https geodacenter asu edu software The license agreement can be found at https geodacenter asu edu license See the PySAL documentation at http pysal geodacenter org 1 5 contents html the Open GeoDa tutorial at https geodacenter asu edu system files geodaworkbook pdf and the e slides at https geodacenter asu edu eslides Relevant sections of these sources are referenced below 3It can also be used for other time stamped events such as diseases 4PySAL can be downloaded at http pysal org 5Both programs can be downloaded at https geodacenter asu edu software 4 b CAST for Mac OSX c CAST for Linux Ubuntu Figure 1 CAST on Windows Mac OSX and Linux Ubuntu CAST GeoDaNet and GeoDaSpace CAST and these programs can downloaded on the GeoDa Center website for Its linked views for interactive exploratory spatial data anal ysis are similar to OpenGeoDa but OpenGeoDa is designed for the analysis of one area level dataset at a time while CAST can display multiple shapefiles simultaneously as layers in one view and or as separa
18. provide an overview of how to use the main functionality of the Crime Analytics for Space Time program CAST which was developed by Arizona State University s GeoDa Center for Geospatial Analysis and Computation under a 2009 12 cooperative agreement with the National Institute of Justice Since we have made detailed explanations about the methods implemented in CAST available through other GeoDa Center resources this guide does not include these explanations but it references the relevant 2 Overview of CAST The idea is to CAST is designed to detect spatial patterns and trends in crime data make it easy to represent different dimensions and contexts of crime like crime types or neighborhood characteristics in views such as maps graphs and calendars that can be animated over time All of these views are linked to allow analysts to identify how selected subsets of the data such as particular beats are characterized across these dimensions Using statistical significance tests CAST includes several cluster maps and trend graphs to detect where concentrations of crimes are higher or lower than expected The program runs on three operating systems Windows MacOSX and Linux Figure It is designed as a user friendly interface to PySAL the spatial analysis library developed at the GeoDa Center in Python that serves as the code base for its functionality PySAL grew out of Professors Luc Anselin s and Sergio Rey s
19. raphs can be linked to other views like maps as in Figure 15a where the top row shows a map and trend graph for robberies and the bottom row for car theft The selection of three reporting districts is reflected in all four views and demonstrates that the high frequency of robberies in these areas coincides with low car theft incidences The trend lines in the dynamic trend graph are colored according to the colors of the map classification The Play button moves a horizontal yellow line through each time period in the series as the image of the map for the same period changes as well The significant trend graph adds a significance test to a time series of events in a given area to determine subsets in time that are below or above average As for the local cluster maps for areas the two tests that are implemented here are local Moran s I and local G statistics In the trend graph case the test is applied separately to each area Within each area the number of time periods constitutes the unit of analysis The test compares values in one time period to a specified number of previous and or future time periods For instance or 6 month period as found in the sample data of this guide aggregating crimes to 1 week periods provides a sufficiently large number of intervals about 25 a rule of thumb is to have at least 20 periods for this analysis If five past time periods are selected as neighbors Figure the
20. te views Like OpenGeoDa CAST reads so called shapefiles the geographic file format related to ESRI s ArcGIS software the most widely used commercial Geographic Information System GIS 3 Prerequisites To get started with analyzing data in CAST you need a point or polygon shapefile which at a minimum consists of three files with shp shx and dbf extensions If you want to load multiple layers in one view or link shapefiles across views as usual they need to have the same spatial extent The tools in CAST will only work with shapefiles that are currently open in CAST e g if you open a shapefile and close it the shapefile dropdown selection option in the query views might still list shapefiles from memory but they need to be open in order to be mapped or graphed 3 1 Projections load a point and area shapefile in the same view the coordinates of the point shapefile need to be in the same projection as the area shapefile For instance if the points have latitude longitude coordinates such as 122 50022 37 718954 but the area shapefile is based on projected coordinate system e g NAD 1983 State Plane California then the point coordinates need to be converted to projected coordinates in order to be displayed in the same view or linked across views in CAST in this example the projected coordinates would be 5983167 999905 2090431 999985 7 5 At http www geodacenter asu edu software TTo convert lat long
21. ter map Figure highlight with its neighboring areas highlighted in red It also works for dynamic area and select your spatial weights matrix This enables the selection of an area yellow selection brushing Figure 134 gives an example of dynamic LISA map the dynamic local G statistics map has the same interface 6 2 Calendar Maps In calendar maps two monthly calendar views are integrated with a regular CAST map view Figure 14a 1 Micro calendars from January to December and enlarged versions of two months that correspond to a selection of months in the micro calendars The days in the two calendars are color coded based on the number of crimes events that occurred on each day with lighter green colors for smaller frequencies and darker green shading for larger ones In addition the enlarged two month calendar also includes the date and frequency count for each day Finally bar charts on the vertical axis of the calendar summarize the number of crimes for each week while bar charts on the horizontal axis display the sum of crimes for each day of the week In the dynamic version of the calendar map each day in the enlarged lFor more information about spatial weights local cluster map settings and interpre tation see chapters 15 and 16 weights and 19 LISA maps in the OpenGeoDa tuto ril at https geodacenter asu edu system files geodaworkbook pdf and the video tutorials at https geodacenter as
22. u edu eslides For background on the methods see Anselin Luc 1995 Lo cal indicators of spatial association LISA Geographical Analysis 27 93 115 and the free e course about local spatial autocorrelation at http moodle geodacenter org course view php id 7 For information about the local G statistics methods see Getis A and Ord J K 1992 The analysis of spatial association by use of distance statistics Geographical Analysis 24 189206 Ord J K and Getis A 1995 Local spatial autocorrelation statistics Distributional issues and an application Geographical Analysis 27 286306 21 Define neighboring areas based on shared corners borders queen or shared borders rook Define points as neighbors within a given threshold distance or for given number of neighbors c Dynamic LISA Cluster Map Figure 13 LISA Cluster Maps 22 calendars is highlighted sequentially and the corresponding points for this day are displayed on the map Figure 6 3 Trend Graphs To track frequencies of crime over time CAST includes three types of trend graphs 1 a stand alone static trend graph with crime counts on the vertical axis and time periods on the horizontal axis 2 a dynamic trend graph that integrates this stand alone trend graph with a map view and 3 a significant trend graph that highlights statistically significant time runs within the time series associated with an area Static trend g
23. ualize the cluster core The full cluster including neigh bors can be shown for selected areas through the Select Neighbors right click map option since neighboring areas can be part of multiple clusters or represent cores and clusters Sim ilarly a trend line can be a neighbor of several time periods For instance in the example above the week of Dec 11 17 2012 would be a 5 week previous neighbor of both Dec 25 31 and Dec 18 25 and could be significantly related to one but not to the other To display the results of the neighbors specified in the time weights file for a particular time period one can click on the red triangle at the top of the trend graph This will front load the color coding for the neighbors of this time period 27 7 Outlook The alpha release of CAST is currently available for free download on the GeoDa Center s software page Alpha releases are the initial software releases that are still part of the testing phase with active debugging As such we are currently responding to bug reports with new uploaded versions of the software that addressed the reported bugs since analysts sometimes also suggest improvements in the user interfaces there might be slight divergences between the interfaces of the latest version and the screenshots in this guide but the core functionality remains the same Analysts can send bug reports to geodacenter asu edu Releases are announced on the GeoDa Center s Openspace listserv and
24. ve a new dbf file for your area shapefile from 5 It Figure 7 Options to Aggregate Points in Time and Space and Subset Data 13 Incidents and Percentile c Incidents Heatmap and LISA Cluster Map Figure 8 Multiple Layers of Robberies 14 Figure 9 Heatmaps for Four Crime Types surfaces as third layers as shown in Figure 8 Shapefiles can be opened in separate views with the same spatial extent or with different spatial extents selecting subsets of the data will work between shapefiles that share the same spatial extent Figure 9 shows an example of four different crime types in San Francisco one in each view which illustrates that drug arrests are more spatially concentrated than vehicle thefts 5 3 Linking and Brushing Figure I0a illustrates that the linked selection of observations works across views scales and map types P In this case points associated with a heatmap are linked to grids and police districts in a percentile map and a LISA cluster map The figure is based on four shapefiles and their associated tables SF plots drug points vandalism points and grids that all share the same spatial extent When a rectangular subset of this spatial extent is selected in one view the same subset is selected in the other views If the same shapefile is loaded in two views they will share the same table h
25. vem Mop 0000000000003 a Heatmap and Settings for All Time Periods Combined Dynamic Density Mas Tepet90106 083197 Time of Bay 00 00 23 59 10 el m Interval Start 09 01 2006 199 08 31 2007 M mee Celisize 125 4891 34 X 400 image Bandwidth 1421 989 w 29401 7 50195 D Je cens GP Show cumulative density muda 279 Cem b Dynamic Heatmap and Settings PUE Dynamic Heatmap Figure 12 Static and Dynamic Heatmaps 20 significantly higher lower values compared to the overall area F To create these cluster maps one first needs to define what constitutes a neighboring area This is done through a so called spatial weights matrix which in CAST allows analysts to define neighboring areas in terms of shared borders and or corners or neighboring points in terms of distance bands and a given number of closest points Figure shows the options for specifying spatial weights matrices in CAST Note that the local cluster maps display the cores of significant clusters that do not include the neighbors since one area can belong to multiple clusters or represent a core and neighbor To display cores and neighbors in other words the whole cluster right click on the static or dynamic clus
26. y boxplots histograms and bivariate scatterplots Figure B shows the toolbar available for all map views Dynamic map views have an extended toolbar that also includes play pause stop and other typical movie options The toolbar icons will be familiar to many analysts and are labeled in Figure 3 Starting from the left the first option is the Add Layer feature to for instance load a time stamped point shapefile and a polygon shapefile with the same projection which are input files for several tools The only non standard feature is so called brushing i e dynamic selection of observations which can be accessed through the brush icon 10 drugs vandal Change the legend category color Change the outline color e g to make outline less dominant Display areas without outline to make patterns more visible Turn outline back on Figure 5 Map Legend Color Dropdown Options 4 2 Context Specific Dropdown Menus Right clicking on a layer in a map view will result in the display of the dropdown menu shown in Figure d The LISA map choice is one of the cluster maps described below while the map classifications are the same as described above Change transparency is useful when applied to a top layer for displaying additional layers below When right clicking on the color legend of a layer displayed in the map view the dropdown menu shown in Figure provides options to change the legend color and polygon or po
Download Pdf Manuals
Related Search