topic Re: Additive interaction estimates in PROC GLIMMIX in Statistical Procedures

Additive interaction estimates in PROC GLIMMIX

njgrubic — Wed, 18 Nov 2020 17:44:13 GMT

I have created a multi-level logistic regression model using PROC GLIMMIX. I am able to obtain interaction estimates and 95% CIs on the multiplicative scale from the fixed effects output, although I am not sure if there is a simple way in SAS to output the estimates and 95% CIs on the additive scale (I am looking to calculate relative excess risk due to interaction [RERI] with a 95% CI). SAS code below. Thanks.

proc glimmix data=cares2;
class group outcome (ref="No") intervention (ref="No") smoker(ref="No") area(ref="urban") gender(ref="Female") /;
model outcome = intervention area intervention*area gender / dist=binary link=logit ddfm=bw solution oddsratio cl;
random intercept / subject=group;
lsmeans intervention*area / slicedifftype=control('No' 'urban') adjust=bon oddsratio cl ilink;
run;

Re: Additive interaction estimates in PROC GLIMMIX

SASJedi — Fri, 20 Nov 2020 16:24:11 GMT

Moved to Analytical Procedures forum where it should get better visibility

Re: Additive interaction estimates in PROC GLIMMIX

SteveDenham — Mon, 23 Nov 2020 13:49:53 GMT

The ilink option provides estimates on the original (additive) scale. However, to get additive differences to generate relative risks I suspect you will need to use the %NLmeans macro. The note here should provide a starting point.

SteveDenham