forum.smartpls.com

Dear all,
My DVs are binary and my IVs are binary as well as metric. I wanted to inquire if there is any way through which I can run the model. My sample size is bit small so I would prefer to use PLS only. Can anyone please suggest me how to ascertain the constructs reliability and validity in PLS, if my DV is binary.

Thanks and regards,
Akanksha Goel

If your DV is binary the regression techniques that are used by PLS become problematic. It is still possible to run the algorithm and get results, but they would be similar to results from a linear probability model in regression with a binary outcome. Such a model has some downsides over logistic regression, for example, that predictions could be outside of the range of possible values or that estimated standard errors are not reliable when the relationship is none linear, but sigmoidal.