User Manual for PovMap - World Bank Internet Error Page
Contents
1. oOutputFileName 0nnn ConfigFileName PovMap configuration file name v View input parameters of Alpha and Beta A showing Alpha vector showing Beta vector V showing varian covarian matrix L showing linkage vector T showing raw data only for memory mode 1 h Show this help screen E Reset output file Existing file will be erased nN Override the number of simulation defined in config file oOutputFileNam Alternative output file name Default is Result dat 0 Set switches to random drawing 0 suppress random drawing on Beta 0 suppress random drawing on locational effect 0 suppress random drawing of household effect 18 The command line options need further explanation e r Reset the output file By default the output of bootstrapping will always append to the end of existing result file POU To clear the content in the result file use r option e n e 0 Override the number of simulation specified in PCF file Testing switch Is followed by three digits of 1 or 0 When 0 is shown in that position the correspondent random variable will be shut down This is useful when testing the result against an existing measurement e y Requests a display of parameters including estimation of Beta and Alpha its variant covariant matrix Run PovMap under Windows PovMap can also easily run under Windows User can associate the file type PCF with the executable PovMap exe and then dou2. 00101 35 3000 10 4 100102 50 2000 1600 7 5 Please keep in mind that all records containing missing values will be dropped Compounded ID in Census Cluster ID is the ID that identifies the lowest level in survey dataset and is typically the county level It may not be the lowest level in which you want to produce the poverty map measurements Even though PovMap provides aggregations for the cluster level and the levels below going lower than cluster level is typically associated with larger standard error Only one identifier is allowed in the PovMap s PDA dataset in order to save space and to run faster This identifier could be as long as 16 digits and is typically compounded with identification of different administrative units Suppose the dataset survey and census have multiple identifiers such as STRATUM ranged from 1 to 9 COUNTY ranged from 1 to 99 DISTRICT ranged from 1 to 125 and TOWNSHIP ranged from 1 to 999 a compounded ID at district level can be created with DISTID STRATUM 100 COUNTY 1000 DISTRICT 11 or DISTID STRATUM 100000 COUNTY 1000 DISTRICT Similarly a compouneded township ID could be defined as TOWNID STRATUM 100 COUNTY 1000 DISTRICT 1000 TOWNSHIP or TOWNID STRATUM 100000000 COUNTY 1000000 DISTRICT 1000 TOWNSHIP The multiplier e g 100 or 10000 in the a
3. Bound v Optional Specifies the range for household effect trimming The value v could be AUTO NONE or a specified value When hBound A UTO the range will be hBoundInSurvey hBoundInSurvey where hBoundInSurvey is the highest absolute value of household effect in survey When hBound NONE no trimming will be used When v is specified as a number the trimming range will be v v bBound v Optional 0 lt v lt 1 The accepting probability for drawing random vector Betl Value v will be internally converted to range according to the distribution of the random variable aBound v Optional O lt v lt 1 The accepting probability for drawing random vector Alpha model Value v will be internally converted to range according to the distribution of the random variable INDICES Required Indices of poverty and inequality measurements User can choose from FGTO FGT1 FGT2 GEO GEI GE2 ATK2 GINI in any order General Entropy measurements with fraction parameter can be specified like GE1 75 User can also require the distribution to be outputted as percentile Three different styles can be used here DIST DIST 20 or DIST 1 5 10 25 50 75 90 95 99 The first specification produce the distribution as decile value the second format allows for 20 groups of percentile equally spaced The third format will produce an uneven group at 196 596 10 95 99 levels Simulationzn n2 Required This is the way to run simulations on different aggr
4. DUCHD HEADAGE NOWIFE EDUCWF WIFEAGE PPD YGIENE1 YGIENE2 YGIENE3 MATER1 MATER3 COCINA1 COCINA4 VIVI1 VIVI2 NATIVE OPERSON TOPER2 TOPER3 PPD2 PPD3 CMEAN2 CMEAN3 CMEAN25 CMEAN40 SCMN22 arhs VAR4 VAR11 XBETA4 S12 S112 S23 S45 S46 S49 S55 S810 S314 S417 S420 S421 S721 1019 1121 1219 1221 _Yhat_ _Yhat Cluster CLUSTER sWeight FACTORES CenData smallcengabe dat cWeight OPERSON cKeyVar CLUSTER LocErr YES DataOut small The specification is identical to the SAS version All previous explanations for SAS apply here except the following 10 sDir or cDir clause is not needed User can put the path of file in front of the dataset name Le srvdata c project1 stratum3 LSMSGABE4 dta However cOnlyVar is not supported The following dataset formats are supported by PovMapPacker Stata file DTA Any Stata format between version 2 0 to 8 0 is supported However the Stata file generated with an Unix system may not be read by PovMapPacker dBase file DBF DBF III or IV are supported including other variations such as Fox pro s DBF file ASCII file with header This is an ASCII file in which fields are separated by comma missing value is coded with no symbol The following example has a missing value on field INCOME at first record HHID HHAge INCOME EXPEND EDUC HHSIZE 1
5. KK KK KKK KKK KK KKK KKK KKK KK KKK KKK KKK KKK KKK KKK KK KKK KKK KKK KKK x Data preparation for PovMap KKK KKK KERR EK KKK KK RK KKK KKK RK kk RK KK KKK KR KR KKK KKK KR KK KKK KEK RR KK KKK KEK KER KK KKK KEKE RK y include C projects PovMap fileaux sas include C projects PovMap dataprep sas KIlLlclllllLlllIcl Specify the Survey information E dir of survey data Slet sdir input dataset name Slet srvdata LSMSGABE clustering variable Slet Cluster CLUSTER survey weight let sWeight FACTORES RARE LAS varrable let lhs LRPCEXP Beta RHS variables Slet rhs EDUCHD HEADAGE NOWIFE EDUCWF WIFEAGE PPD YGIENE1 TOPER2 TOPER3 PPD2 PPD3 CMEAN2 CMEAN3 CMEAN25 CMEAN40 SCMN22 Alpha RHS vars Slet arhs VARA VAR11 XBETA4 S12 S112 S23 S45 S46 Locational effect S let LOCERR YES Specify the census information E dir of Census data Slet cdir input dataset name let cendata smallcengabe census weight let cWeight OPERSON ID Vars in census let cKeyVar Cluster Cluster only vars let cOnlyVar CMEAN2 CMEAN3 CMEAN25 CMEAN40 SCMN22 Cluster only file S let cOnlyDat cclusterOnly A Output Directory H out
6. LS in model 2 so a GLS regression is needed In GLS the variance covariance matrix is a diagonal block matrix with structure 6 tO o o e On O o e e o On O e 8 c c e Opto Overall the procedure for stage 1 of of the poverty mapping computation can be listed as sl estimate Beta model 2 s2 calculate the location effect 3 2 s3 calculate the variance estimator var o 4 s4 prepare the residual term En for estimating Alpha model 6 s5 estimate GLS model with 8 s6 use a singular value decomposition to break down the variance covariance matrix from previous step This will be used for generating a vector of a normally distributed random variable such that the joint variance covariance matrix will be in the form of 8 s7 read in census data eliminate records containing missing values generate all census variables needed for both Beta and Alpha models s8 save all datasets needed for the simulation the PDA file Dataset and Memory Usage The dataset generated during the data preparation stage is in a proprietary format with an extension of PDA The dataset includes the regression results from the Beta and Alpha models decomposed variance covariance matrix from step s6 decomposed variance covariance matrix from step s4 household count of each cluster and other parameters estimated in the data preparation step and finally the census data in binary format The goal is to o
7. User Manual for PovMap Version 1 la Qinghua Zhao Development Research Group The World Bank 1818 H Street N W Washington D C 20433 Introduction PovMap is a software package that computes poverty and inequality indicators at a spatially disaggregated level Poverty mapping is a method that uses a model of household expenditure model from a survey dataset to estimate household welfare in a census dataset which typically do not include household expenditure or income information Poverty indicators at the community level are then formed as aggregats Bootstrapping is used to improve the accuracy of the estimation The method consists of two stages During the first stage a series regression is run to model the expenditure and decompose the random unexplained component In order to apply this to the census data the regressors in the model need to exist both in the survey and census datasets A special dataset that combine census data and model parameters is then produced The second stage of poverty mapping is the simulation stage This stage uses the model parameters but performs repeated drawings on different random components to bootstrap the household expenditure The basic structure of this package is shown in the chart below The data packer estimates the model and organizes the information for simulation Two packers are provided for the user one that uses SAS software and the other that does not SAS users will use a se
8. X become 0 99 X 0 02 lt gt assign X 0 02 to X lt gt X become 1 02 X 0 02 lt gt assign X 1 02 to X lt gt X become 0 98 An number can also be assigned to the Beta with Beta VarName 1 23 e Alpha VARI1 0 Alter the value of Alpha model Similar to Beta e seed last Set the random seed to be the same as last simulation e yBound 0 999 Another way to trim Y Specify the proportion to be kept 16 Identify the Distribution of Random Component An earlier version of this program included a module that identified the distribution of random components Unfortunately this module does not work well In the current version this function is excluded User have to do analysis manually to identify the best fitted distribution For SAS user the data preparation will produce two SAS datasets with name ClusterRes and hhldRes ClusterRes is the random component correspondent to cluster effect 7 hhldRes is the random component correspondent to idiosyncratic component en For user of PovMapPacker the two files produced have extension pResC and pResH and both are in ASCII format Obs Weight Residual 1 3 2776626e 003 0 17330292 2 4 8687609e 003 6 4200165e 002 3 2 2275377e 003 2051137692 4 1 3126561e 003 0 39374108 5 3 5004163e 004 4 3999818e 002 6 2 0684278e 003 0 275217 7 1 4850251e 004 0 547977156 8 2 1320718e 003 0 14737068 9 Format of yDump file User can collect all the estimated Y into a binary
9. ble lick over PCF file to run PovMap exe To do that open windows explorer right click over the PCF file then click the Open with Under the Open With menu check the Always use this program to open these files then click Other to location the PovMap exe file Click OK once the PovMap exe is identified This will permanently associate PCF file with PovMap exe Result of Simulation Result of bootstrapping is stored in file of type POU The result file is a tab delimited text file The first nine columns are always provided no matter whether you choose the measurement Type Unit nHHLDs nIndividuals nSim Min Y Max Y avg MEAN se MEAN Point Estimate 3227 145 760 100 5698 7599 1628399 4 94299 803 12172 398 Point Estimate 3228 120 605 100 7082 1178 1253250 2 85392 196 10001 359 Point Estimate 3229 134 782 100 8621 7361 1034800 7 74580 428 10317 259 Point Estimate 3230 143 802 100 8850 0348 2258822 8 85437 389 9125 4147 Point Estimate 3231 112 634 100 2629 6289 611199 37 43242 205 4512 5221 Point Estimate 3232 59 382 100 2134 0325 302344 98 31155 461 2907 7902 Point Estimate 3233 57 288 100 5045 925 711865 91 57055 221 7955 2547 19 The remaining columns are optional depending on whether it is specified in INDICES statement of PCF file Each index selected will have two columns the first one denoted with prefix avg is the ave
10. bove example should be carefully chosen according to the range of each identication When users want to estimate the poverty and inequality at levels lower than cluster level they should use both CLUSTER and CKEYVAR clauses such as Slet Cluster DISTID let cKeyVar TOWNID Then the cluster ID has the form of SCCDDD and the ID on each record of census data has the form of SCCDDDTTT During the simulation cluster ID will be used to determine when a new cluster level redraw should happen this ID will be used to produce aggregations at different levels by shifting the ID to the right Stage 2 Bootstrap Simulation Methodology The fully specified simulation model is defined as follows 9 Inj 2x B Ft o lt 2 where p N B 5 ij is a random variable could be normally distributed or T distributed with a variance defined in 5 n is a random variable either normally distributed or T distributed with a variance defined in 7 B exp ZL amp and amp N Trimming could be applied to the random variable 77 and en as well as to random vector p 12 and In the case of a normal distributed random variable a range 1 96 1 96 will make 10 of random N 0 1 drawing to be redrawn For random vector of size m the vector will be redrawn if the mode of the vector a 7 distributed random variable is outside the specified range After estimating In y several poverty and ineq
11. d to as Beta model since survey data is just a sub sample of the whole population the location information is not available for all regions in the census data Thus we cannot include the location variable in the survey model Thus the residual of 2 must contains the location variance 3 Uon 7 is u Here is the cluster component and En ig the household component As mentioned above the estimate of e for each cluster in the census dataset is not applicable therefore we must estimate the deviation of Taking the arithmetic expectation of 3 over cluster c 4 Hu 1 Hence gt Efu 0 var g o TT Assuming Me and ch are normally distributed and independent each other Elbers et al gave a estimate of variance of the distribution of the locational effect 1 232 Kiel 5 var 02 X a var u b var 72 Y 2 a2 02 72 y 20272 pg n 1 When the location effect 1l does not exist equation 3 is reduced to Uon Eon According to Elbers et al the remaining residual 4 can be fitted with a logistic model and will regress a transformed En on household characteristics 6 i ch share 7 A t ch also referred to as Alpha model 2 where A set to equal 1 05 max Ech The variance estimator for En can be solved as AH i Liri l H1 I L dd Hy 7 The result from above indicates a violation of assumptions for using the O
12. egate levels When n 0 the record identifier in PDA file will be used When this ID changes a new aggregation will be outputed When n gt 0 the ID in census dataset will be shifted n digits to the right to produce a shorter ID that represents an aggregation on higher level new aggregation will be outputed when this value changes For example if ID is the form SCCDDD then SIMULATIONS will produce a estimates at the county level SCC 15 When multi level simulation is requested such as Simulation 0 3 5 estimates of district level SCCDDD county level SCC and stratum level S will be produced in one simulation Please note that characters used in simulation configuration file are not case sensitive Users may also use in front of each line to disable that line set it as a comment line Other Options All items listed here are optional e END To terminate the execution Any statement after END will be discarded e yDump v In order to dump the estimated Y This is useful when further measurement is wanted The output file is binary See below for more information e Beta VarName 1 01 To manually alter the value of Beta model VarName identify the variable name whose value will be set to 1 higher Similar notation could be and Their impact can be shown with concrete example let X 1 0 X 1 01 lt gt assign X 1 01 to X lt gt X become 1 01 X 1 01 lt gt assign X 1 01 to X lt gt
13. equired if cluster effect is modeled Type of distribution for cluster effect Value selected from T n T with DF of n N normal NP non parametric HNP hierarchical non parametric HDist v Required Type of distribution for cluster effect Similar to CDist MinImpute v Optional Lower boundary for trimming simulated LHS variable v could be a numeric value AUTO or NONE When AUTO is used the lower bound of per capita expenditure in survey dataset will be used to eliminate that household in that simulation MaxImpute v Optional Upper boundary for trimming simulated LHS variable v could be a numeric value AUTO or NONE When AUTO is used the upper bound of per capita expenditure in survey dataset will be used to eliminate that household in that simulation Default is NONE cBound v Optional Specifies the range for cluster effect location effect trimming The value v could be AUTO NONE or a specified value When cBound A UTO the range will be ScBound ScBound where ScBound 1s the highest absolute value of 14 cluster effect in survey When v is specified as a number the trimming range will be v v When random number is outside of this range repeat drawing will occur until a random value is inside the boundary When cBound NONE no trimming will be used in fact it is done by setting up a boundary from negative infinity to positive infinity The other boundary setting described below are implemented in this way too h
14. file The option YDUMP The yDump file is organized as double for cluster ID float for household size float for y11 float for y double for cluster ID float for household size float for yz float for yz double for cluster ID float for household size float for y3 float for ya double for cluster ID float for household size float for y 1 float for Yno float for Y1 nSim float for Y2 usim float for ya nSim float for y nSim 17 This file can be read with SAS with following code Data YDUMP infile c project1 stratuml ydump bin recfm n input ClusterID rb8 0 hhsize x1 x100 float4 0 here X100 will be replaced by X300 if 300 simulation were run recfmzn along with read the binary data as a stream User who wants to read the YDUMP file in Stata please contact me Run PovMap under Command Mode PovMap exe can be run in Windows command mode Users can keep PovMap EXE anywhere on the PC or on a network drive For simplicity let us assume the program is placed on the network drive s PovMap and user has already change directory to c PovMap Stratum1 and prepared a simulation configuration file SsTRATUM1 PCF To begin running PovMap type the following c PovMap Stratum1 gt s povMap PovMap STRATUMI PCF For the complete list of PovMap s syntax type option n in command line PovMap h The syntax will be displayed as PovMap ConfigFileName vABVLT h s r
15. icantly increase its size This dataset can be read in from another census cluster mean file with this option let OutDir OutputDirectory to specify a directory to store the output datasets let DataOut OutputPovmapDataFile the output dataset will have extension PDA let LocErrzYesOrNo to specify whether the locational random component should be modeled When equal to NO the location effect will not be modeled Dataprop the last statement to execute the SAS macro DataPrep When editing this file the case of variable names or reserved words are not important and neither is their order except the statement DataPrep which must be the last Please make sure that the name YHAT must not exist in the survey dataset because it is reserved to represent the estimated per capita expenditure Names start with _Z and followed by a number should also be avoided because they are internal variables for the Alpha regression Data Preparation without SAS When SAS is not available or the census dataset can be processed by another package PovMap supplies PovMapPacker exe to prepare the dataset PovMapPacker performs the same task as the version with SAS Currently three types of datasets can be used Stata any version dBase III and IV and ASCII with header line To use PovMapPacker the user needs to prepare a model specification file which looks like KKKKK Model KKKKK srvdata LSMSGABF4 dta lhs LRPCEXP rhs E
16. of the variable that holds the per capita expenditure figure in logarithm let RHS List of Variables in Beta model to specify the variable names in Beta model The variables must exist in the SAS dataset let aRHSzList of Variables in Alpha model to specify regressors in Alpha model Regressors can be either a variable that already exists or an expression that uses variables in the survey dataset Variable name Yhat can be used in the formula to represent the predicted value of Beta model let CDIR CensusDirectory to specify the directory where the census data resides Full path name is allowed Omit the path name if the census dataset is in the same directory as this code let CenData CensusDataset to specify the name of census dataset let cWeight NameOfWeighting VariableInCensus to specify the weighting variable in census data Must exist already let cKeyVarNameOfIDinCensus If the cluster ID variable already exists in the census dataset place its name here This can only be used when the cluster ID in the census is different from the cluster ID in the survey cKeyVar need not to exist in survey dataset let COnlyVar Variables Optional It defines the variables at cluster level and does not vary within a cluster Variables of this type will be stored separately and restored during simulation let cOnlyDat DatasetNameWithCOnlyVar cOnlyVar as defined above need not exist in the census data because it will signif
17. put directory let outdir let LOCERR Yes Slet dataout small dataprep The above program is written in SAS macro language In this SAS code statements starting with and end with are a comment statements and have no programming meaning SAS will ignore all comment lines The meanings of other statements are explained below include statement will retrieve an external SAS code from a specified file Two files to be included are FileAux sas and DataPrep sas Full path specification is allowed If the user places these two file into SAS s macro directory these two include statements can be omitted let SDIR SurveyDirectory specifies the directory of the survey dataset If the survey dataset is in the same directory leave it empty between and Please note that no quote sign is needed even if there are spaces in the directory name i e sDir c temp will cause problems but sDir c temp is correct let SrvData SurveyDatasetName to specify the name of survey dataset in SAS format Must be supplied let Cluster NameOfClusterID to specify the name of cluster ID This variable should exist in both the survey and census datasets and have the same name let sWeight NameOfWeighting VariableInSurvey to specify the name of the weighting variable within the survey typically compounded by household size and cluster weight let LHS PerCapitaExpenditureInLogrithm to specify the name
18. rage and the second denoted with prefix se is the standard error The best way to open POU file is to associate the POU file with Excel Remark This software package is provided free of charge PovMap is intended for use by the World Bank and its clients and is not intended to be sold or used for commercial purposes Under no circumstances shall The World Bank be liable for any loss damage liability or expense incurred or suffered which is claimed to result from use of PovMap 20
19. rganize all intermediate results into one file which would then take up less hard disk space Even though users may never need to see the binary dataset directly it is still necessary to explain the structure of this dataset Since we are dealing with census data the program should not be limited by the size of this data For the sake of computing speed the data will be read into memory however this is always limited Thus it is important to compress the data efficiently In this package a bit type variable is provided implying this variable will occupy only one bit Since there are typically a lot of dummy variables within the census data using the bit data type will allow us to compress up to 8 dummy variables in to one single byte To further conserve space the cluster ID is not stored as a column since they are constant within the cluster PovMap may pack the cluster level information separately to save space Cluster only variables are those that are invariant within a cluster and do not need to be repeated in the household record In PovMap exe cluster only variables will be added to household record The savings associated with using cluster only variables is tremendous For example the PovMap dataset for South Africa stratum 7 with 1 8 million household and 10 variables is barely 25M In order to achieve the fastest speed and most efficient use of memory four memory usage modes were designed into the PovMap exe simulator The
20. t of SAS programs to handle all tasks in stage 1 Others will use a stand alone executable PovMapPacker exe to prepare their data survey i Simulation 1 Configuration Step 2 PovMap exe MIN census j Step 1 Packed i Data Prep i census B i A Model zc pc CM M MM C CREE nObs nVars Specification nObs nIRecL Whay this design SAS is widely used to perform data processing of census data SAS has all the necessary modules built in to perform tasks such as GLS and singular decomposition Without SAS software it is difficult to read SAS datasets SAS users must perform the data packing within SAS Users who do not have access to SAS must utilize other statistical packages for model building Presumably their dataset is much smaller The stand alone data packer PovMapPacker exe can access Stata DBF and ASCII formats The packed dataset between stage 1 and stage 2 can be made more compact to allow the dataset to be fully loaded into memory This option is much quicker than throughdisk operation A stand alone EXE executable can efficiently satisfy these demanding computations It is necessary for the poverty mapping software to run efficiently during the simulation s
21. tage to ensure its success Our version of PovMap requires at most a few minutes to run When executing the PovMap software most of the computer s system resources will be occupied however the user will still be able to perform other tasks and switch between programs PovMap can also be run in conjunction with other processes that are currently reading the same data file and outputting the result to that same file This package is designed to be used by the general user Users do not need to be able to program in SAS as long as an easy specification file can be made Computing Model In this section the computing process for poverty mapping will be summarized Users of this manual should always refer to the paper by Elbers Lanjouw and Lanjouw 2001 for theoretical background and statistical inference The computing of poverty mapping begins during the estimation of the expenditure function For simplicity we assume per capita expenditure of a household is the basic left hand side variable and the word cluster is an aggregation level in survey and census datasets In Yon Elin ya X4 u 1 where c is the subscript for the cluster h is the subscript for the household within cluster c Yeh is the per capita expenditure of household 7 in cluster c Xen is the household characteristics for household in cluster c a linear approximation of model 1 is then written as 2 IN Yo Xen B Hey also referre
22. uality measures will be computed They include Generalized Entropy class A GE A zz d Az0 Axl 1 y 1 y y GE 0 w logg and GE lD w log W 2 y W gt yY y 1 c Kae 1 Atkinson class of measures A c 21 W H 0 l y and Gini index W il 2 W 1 WW Dy Gini wy Lp 0 5 w 5 where Dia Pi tw In the above definitions w is the weight of household i and W is the total population Simulation Configuration To run the simulation the user needs to create a configuration file with extension PCF DataSource Stral pda nSim 100 PovLine 45476 MemorySize 128 SEED 1234567 CDist T 5 HDist N INDICES FGTO FGT1 FGT2 GEO GE1 GE2 ATK2 GINI 13 minimpute none MAXimpute none HBound auto CBound auto ABound NONE BBound 0 99 Simulation 0 3 6 Explanation DataSource Optional The filename of the input PDA file If omitted a PDA file with same name as PCF is assumed NSim n Required Number of simulation PovLine n Required Poverty line MemorySize n Optional Memory size available to the simulation Default is 128M Seed n Optional The user specifies seed for random number generator on Beta vector When omitted or set to 0 internally produced random seed will be used This seed is derived by the system clock in 1 1000 second resolution Thus no two simulations will be equal if seed is set to 0 CDist v R
23. y will be selected by automatically program depending on the available memory set by the user These four modes are Mode I data is internally organized as a matrix of X and Z in double precision Demands n 8 m t mp 2 bytes of memory When n 1 000 000 m 20 and m 10 PovMap needs at least 256 Mega bytes of memory This is the fastest mode Mode II demands much less memory as it needs only n m 16 bytes of memory Here m is the record length of census data If m 260 and n 1 million PovMap needs 76 Mega bytes of memory space This mode is about 1 5 times slower than the mode 1 Mode III the highest memory saver but the slowest mode PovMap needs only n 16 bytes of memory Data will be read directly from the census data and processed on the fly The speed of mode III is dependent on the computer s throughput Thus when a larger dataset is used it is better to process it on the local drive Stage 1 Preparing Data As previously mentioned there are two packages users can select from when preparing a PovMap dataset SAS allows users to perform data preparation and process census data The necessary components are SAS Base SAS IML and SAS Stat For those who do not use SAS for census preparation we also provide a tool that can read datasets in Stata dBase and ASCII formats Data Preparation with SAS Users do not to be SAS experts as long as the model can be specified into the following format KKK KKK KKK KKK KK KKK KKK KKK K
Download Pdf Manuals
Related Search