topic Re: proc causalmed: How to avoid "Percentage Due to Interaction"? in Statistical Procedures

proc causalmed: How to avoid "Percentage Due to Interaction"?

YifeiRen — Mon, 17 Jan 2022 21:28:27 GMT

proc causalmed data=b all;
class dementia pegender edugroup cluster5/ref=first;
model dementia = gdsf cluster5 ;
mediator gdsf = cluster5;
covar age pegender edugroup;
Bootstrap NBoot=5000;
run;

I used the code as showed above. But the results always showed with the "Percentage Due to Interaction" although I didn't add an interaction variety in the model. Can anyone help me, please?

Re: How can I run a result without "Percentage Due to Interaction" in proc causalm

ChrisNZ — Mon, 17 Jan 2022 21:27:15 GMT

Retitled and moved to STAT.

Re: proc causalmed: How to avoid "Percentage Due to Interaction"?

SAS_Rob — Tue, 18 Jan 2022 11:52:09 GMT

For binary responses, you will still get a nonzero proportion due to interaction even when you do not fit the interaction. This is because the model is non-linear; the effect attributable to interaction is on the additive difference scale, but the model is being fit on the log or logistic scale. This is a known and derivable quality of Causal mediation analyses, but it is not clear how you would actually interpret when you don't fit the interaction explicitly.