I have a piecewise growth model w/3 parameters (i, s1, s2). Want to see whether growth parameters predict a distinct binary outcome (smoking abstinence). I have run 2 ways: a) Using predicted factor scores from growth model output as IV's in a logistic regression in a separate model, b) regressing the binary outcome on int and slope factors within the growth model itself. However, I get diff results.
Can you explain why? I suspect has to do with int/slopes as latent versus observed variables. How would you proceed?