How to manually derive posterior latent class probabilities with finite mixture model

[Andrew Xavier]

Dear all,

I have the following code

clear
webuse bc
stset t, fail(dead)
fmm, lcprob(smoking): (streg smoking, dist(exponential)) (streg smoking, dist(exponential))

fmm, coeflegend

predict probabilities*, classposteriorpr

gen predicted_class=1 if probabilities1>=0.50
replace predicted_class=2 if probabilities1<0.50

tab predicted_class

The code above is used to predict which class we expect an individual to belong.

Let's say that instead I have the model estimates and want to compute posterior latent class probabilities for an external dataset. How can I compute this? Essentially how do I derive manually the probabilities obtained from the command "predict probabilities*, classposteriorpr" for the command above?

Many thanks

Andrew