Regression when holding out k cases
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 | * (Q) Do you know of anyone that has written syntax to hold out more than one case and do the cross-validation on x number of held-out cases; I know the adjusted predicted variables will hold out one case, but it would be nice to hold out, say, 5 cases to get a more representative assessment of the regression skill... * (A) rlevesque@videotron.ca 2001/09/01. * http://pages.infinit.net/rlevesqu/index.htm SET MPRINT=no. *///////////////////////////. DEFINE !regres( use=!TOKENS(1) /nbcases=!TOKENS(1) /dep=!TOKENS(1) /indep=!CMDEND) COMPUTE caseid=$CASENUM. /* Calculate the number of regression passes = nbcases - use*/ !LET !nbloop=!LEN(!SUBSTR(!BLANK(!nbcases),!use)) !DO !casebeg=1 !TO !nbloop /* compute caseid of last case used to compute coefficients */. !LET !caseend = !LEN(!SUBSTR(!CONCAT( !BLANK(!casebeg),!BLANK(!use)),2)) /* compute the select variable*/. COMPUTE !CONCAT('svar',!casebeg)=0. IF RANGE(caseid ,!casebeg, !caseend ) !CONCAT('svar',!casebeg)=1. REGRESSION /SELECT= !CONCAT('svar',!casebeg) EQ 1 /MISSING LISTWISE /STATISTICS COEFF OUTS R ANOVA /CRITERIA=PIN(.05) POUT(.10) /NOORIGIN /DEPENDENT !dep /METHOD=ENTER !indep /SAVE PRED. !DOEND !ENDDEFINE. *///////////////////////////. SET MPRINT=yes. ****** Example 1. GET FILE='c:\\Program Files\\SPSS\\Employee data.sav'. N OF CASES 20. * Say we want to use 15 cases to do the regression (holdout 5 cases). !regres use=15 nbcases=20 dep=salary indep=educ jobtime. ****** Example 2. GET FILE='c:\\Program Files\\SPSS\\Employee data.sav'. N OF CASES 22. !regres use=16 nbcases=22 dep=salary indep=educ jobtime . |
Related pages
...