Dear all,
I am working on a project aiming to check whether the association between a continuous exposure x and a binary outcome y is different across categories of a binary variable z. In other words, whether outcome y occurs at a different value of x for z1 and z2. The variable x is modeled as a restricted cubic spline. I found a paper that tackles the problem with the partial effect approach (Saraf, Kidney360 2022, attached).
For my model, I use the mfxrcspline approach from the postrcspline user-written program and I'm able to replicate the curves above. My question is: how would you check that the threshold of x at which the association becomes significant in the groups (in the example below, 34/85/52 respectively) is statistically significantly different across groups?
Many thanks,
Manuel
I am working on a project aiming to check whether the association between a continuous exposure x and a binary outcome y is different across categories of a binary variable z. In other words, whether outcome y occurs at a different value of x for z1 and z2. The variable x is modeled as a restricted cubic spline. I found a paper that tackles the problem with the partial effect approach (Saraf, Kidney360 2022, attached).
For my model, I use the mfxrcspline approach from the postrcspline user-written program and I'm able to replicate the curves above. My question is: how would you check that the threshold of x at which the association becomes significant in the groups (in the example below, 34/85/52 respectively) is statistically significantly different across groups?
Many thanks,
Manuel
