Hi, I'm working on a GMM and I would like one of my classes to be a class of zero use. I'm working with a continuous variable of cigarettes smoked in the past 30 days. Of course there is a preponderance of zeros. I'm not using ZIP as the Mplus suggestion was to not do sue do to some large values. What is the best strategy to assure one class of absolutely no use at all (i.e., not smoking any cigarettes in th past 30 days) as it's not occurring on its own.
With categorical indicators, this is fairly easy because the thresholds can be fixed at +15 in one class. I'm not sure how to do this in the continuous case. You could use two-part modeling and then fix things for the binary part in one class.