* (Q) Would anyone have a bit of syntax that would generate, say 1000 variables, each of which is a randomisation of a variable in the current file? My file contains 4 variables. 3 identifiers and one count variable "countvar". I need to generate 1000 randomisations of countvar. * (A) By rlevesque@videotron.ca to SPSSX-L list on 2001/08/13. * http://www.spsstools.net SET MPRINT=no. *///////////////////////////. DEFINE !random (id=!TOKENS(1) /myvar=!TOKENS(1) /nb=!TOKENS(1) /others=!CMDEND) /* id must be a numeric variable, without duplicates.*/ /* myvar the variable to be randomized.*/ /* nb the number of variables containing "myvar values" in the final file*/ /* others the other variables to be kept in addition to myvar and id */ LOOP cnt=1 TO !nb. XSAVE OUTFILE='c:\\temp\\temp.sav' /KEEP=cnt !id !myvar !others. END LOOP. EXECUTE. GET FILE='c:\\temp\\temp.sav'. COMPUTE draw=UNIFORM(1). RANK VARIABLES=draw BY cnt /RANK INTO rdraw. SORT CASES BY rdraw cnt. DO IF $CASENUM>1 & rdraw=LAG(rdraw). COMPUTE id=LAG(id). END IF. VECTOR v(!nb F8.0). COMPUTE v(cnt)=!myvar. AGGREGATE /OUTFILE=* /BREAK=id /!others=FIRST(!others) /v1 TO !CONCAT('v',!nb) = FIRST(v1 TO !CONCAT('v',!nb)). * Next command is just to show that the variables have the same sum and stddev. * They each contain the same values but in different order. SUMMARIZE /TABLES=v1 TO !CONCAT('v',!nb) /FORMAT=NOLIST TOTAL /TITLE='Case Summaries' /MISSING=VARIABLE /CELLS=COUNT SUM STDDEV . !ENDDEFINE. *///////////////////////////. SET MPRINT=yes. * Example 1. GET FILE='C:\\Program Files\\SPSS\\Employee data.sav' /KEEP=id gender bdate salary. !random id=id nb=15 myvar=salary others=gender bdate. *Example 2. GET FILE='C:\\Program Files\\SPSS\\World95.sav' /KEEP=country density populatn. COMPUTE id=$CASENUM. !random id=id nb=10 myvar=populatn others=country density.