R McDowell posted on Wednesday, August 17, 2016 - 9:08 am
Hello. I wish to explore fitting a censor-inflated normal model to clustered data in which there is a random slope for predictor A of outcome Y, and for which there appear to be an excess of 0 values in Y than would otherwise be expected.
Before inflation I would specify (for censored outcome Y, censored below):
%within% slope1 | Y on A; slope1 on B C;
%between% Y slope1; Y with slope1;
Can you advise whether a censored-inflated model makes conceptual sense? Is it necessary to have a random slope for predictor A in the predictor of the inflation part of the model, or could it simply be a fixed effect?