I want to allow the effects of a large number of continuous predictor variables to be different depending on which treatment group the individual is a part of.

If I had only three continuous predictor variables (X1, X2, X3), I might fit a linear model like this.

Response ~ Treatment + X1 + X2 + X3 + Treatment*X1, Treatment*X2, Treatment*X3

In this case however, I have a large number of correlated continuous predictor variables. I have used PLS before for problems with correlated predictor variables, but I don't know if PLS can handle interaction terms, or if there is a different "flavor" of PLS for problems like this. (PLS-SEM? Latent Class Analysis? I don't really understand these methods...)

I have considered fitting a PLS model for the subset of the data for each Treatment group, but this doesn't seem like good stats.

I'm also not sure how I will interpret the results, because I have typically relied on effects plots to manually judge the nature of fit interactions, which will be difficult to do with a latent variable approach. Many thanks.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.