07-19-2015 07:06 PM
I am conducting a survey design based analysis using SURVEYLOGISTIC. I am attempting to obtain standardized Pearsons and standardized deviance residuals to conduct model diagnostics. However, it appears SURVEYLOGISTIC doesn't have these features. Alternatively, I have tried running the model using PROC LOGISTIC and the standardized residuals range between -2 to +16 (that is not good but I am not sure doing diagnostics using PROC LOGISTIC is appropriate). I would like to check if I would get the same standardized results using SURVEYLOGISTIC. Any suggestions or help would be great. Thank you.
proc surveylogistic data = temp;
class Y(ref="NO") X1(ref=FIRST) X2 (ref= FIRST) /PARAM=GLM;
model Y = X1 X2 /link=logit ;
output out=out p=predicted stdresdev=stdresmodfs stdreschi=s;
ERROR 22-322: Syntax error, expecting one of the following: ;, /, CUMPRED, L,
LOWER, OUT, P, PRED, PREDICTED, PREDPROBS, PROB, STDXBETA, U,
07-20-2015 09:25 AM
Looks like a post processing approach is needed. If the output dataset includes PRED, XBETA and STDXBETA you can calculate the standardized deviance residua asl (observed - PRED)/STDXBETA, and work from there I think.