using strtification to look at effect modification

[Yara Hassan]

Dear all,
I have read several articles on effect modification and my understanding is that we use stratification to check effect modification in Stata.
my question is can I use stratum-specific analysis to check effect modification if my exposure is not binary?

Basically I'm checking if Age is effect modifer on the relationship between exposure ( control=0, non-symptomatic case= 1, symptomatic case=2) and response (adverse event yes=1, no =0).
Can I do that while my exposure has 3 categories rather than 2 only?

If not, then what is the best way to check if effect medication exist?


Many thanks for your time in advance.

[Maarten Buis]

Typically, you would include an interaction effects, see help fvvarlist . This is more parsimonious and thus tends to have higher power then estimating models on different sub-samples in case you have other control variables. Also including continuous interaction effect (like with age) is easy.