topic Re: Rstudent option in Proc Reg in SAS for dichotomous dependent variable in Statistical Procedures

Rstudent option in Proc Reg in SAS for dichotomous dependent variable

abhim701 — Wed, 16 Nov 2016 17:33:42 GMT

I have a data set with dichotomous dependent variable. Can I use Rstudent option in proc reg to identify outliers for a dichotomous dependent variable?

proc reg data = work.import;
model Target= x;
output out = outdata rstudent= r;
run;

Re: Rstudent option in Proc Reg in SAS for dichotomous dependent variable

Reeza — Wed, 16 Nov 2016 18:56:38 GMT

Couldn't you look at a histogram or distribution of the variable by the outcome variable?

I don't think the R student statistics is designed for categorical variables.

Re: Rstudent option in Proc Reg in SAS for dichotomous dependent variable

Ksharp — Thu, 17 Nov 2016 02:50:35 GMT


No. You pick up wrong model for dichotomous variable.
Check PROC LOGISTIC and OBSSTAT option.

Re: Rstudent option in Proc Reg in SAS for dichotomous dependent variable

abhim701 — Fri, 18 Nov 2016 07:22:11 GMT

Hi,

Could you explain the OBSSTAT option with an example. It would be of great help.

Thanks

Abhishek

Re: Rstudent option in Proc Reg in SAS for dichotomous dependent variable

Reeza — Fri, 18 Nov 2016 07:52:55 GMT

Look at the INFLUENCE option with proc LOGISTIC on the model statement. The documentation has an example on regression diagnostics including detecting influential observations.

http://support.sas.com/documentation/cdl/en/statug/68162/HTML/default/viewer.htm#statug_logistic_syntax22.htm

http://support.sas.com/documentation/cdl/en/statug/68162/HTML/default/viewer.htm#statug_logistic_examples06.htm

If using proc genmod then obsstat is an option. Both will help with detecting outliers.

Re: Rstudent option in Proc Reg in SAS for dichotomous dependent variable

Ksharp — Fri, 18 Nov 2016 08:08:23 GMT

Sorry. Check INFLUENCE of model statement.

model ......./ influence

especially check DIFDEV DIFSQUAR


INFLUENCE< (STDRES) >
displays diagnostic measures for identifying influential observations in the case of a binary response
model. For each observation, the INFLUENCE option displays the case number (which is the sequence
number of the observation), the values of the explanatory variables included in the final model, and
the regression diagnostic measures developed by Pregibon (1981). The STDRES option includes
standardized and likelihood residuals in the display.