topic Re: Is it necessary to test for outliers/influential observations in GLIMMX (binomial, logit)? in Statistical Procedures

Is it necessary to test for outliers/influential observations in GLIMMX (binomial, logit)?

marcel — Sat, 23 Sep 2017 07:40:22 GMT

Hello,

This note http://support.sas.com/kb/32/471.html indicates how to test the assumption in a GLIMMIX model, but it does not mentioin how to check for ourliers or influential values. Does this mean checking for outlies is not necessary in GLIMMIX?

This example https://stats.idre.ucla.edu/sas/seminars/sas-logistic/proc-logistic-and-logistic-regression-models/ shows how to evaluate Influence Statistics in sas PROC logistic. I assume it is necessary to check for outliers in GLIMMIX then.

Regards,

marcel

Re: Is it necessary to test for outliers/influential observations in GLIMMX (binomial, logit)?

PGStats — Sat, 23 Sep 2017 19:13:16 GMT

Only robust procedures give you some automatic protection against outliers. Elsewhere you should always be on the lookout for them.

Re: Is it necessary to test for outliers/influential observations in GLIMMX (binomial, logit)?

marcel — Sun, 24 Sep 2017 00:16:14 GMT

Mr. PGStats,

I thought so. Thank you so much for the clarification.

Regards,

Marcel