I'm dealing with a cross sectional health survey and I want to test if a Z variable (three-levels categorical variable) is an effect modifier of the association between X (binary exposure) and Y (binary outcome) considering some confounders (another variables in the model). I would like also to compare the intensity of the effects for each strata of Z.
I was thinking about a powerfull aproach to do that and it would be:
1.Test the significance of interaction between X and Z (XZ) in a Poisson Model (because I'm interstead in a Incedence rate ratio - IRR) controled by confounders to check if there is an effect modification. 2.Evaluate the marginal effect of XZ (and respectives 95% CI) to understand how the effect varies by Z.
Would you suggest another approach for 2? I'm not confident if this is a proper approach to my question.
