Daniel posted on Wednesday, September 01, 2004 - 7:18 am
Hi, I read the Mutehn chapter 19 "Latent Variable Analysis," in the Kaplan textbook (I think it's Kaplan?), and conducted my revision as par suggestions. Now, because I regressed my binary distal outcome on covariates, I no longer get the automatic odds ratios I got prior to this. Is there a way to make this comparison in the program, or do I go back to the old way of comparing thresholds, taking the difference etc???
bmuthen posted on Wednesday, September 01, 2004 - 10:53 am
The ORs are not given with x's since they could differ as a function of x. So you would have to go back and compute them by hand.
Daniel posted on Friday, September 03, 2004 - 10:32 am
what specifications would I use from tech 1 to find the variances and covariances?
bmuthen posted on Friday, September 03, 2004 - 10:34 am
Variance and covariance specifications show up in the Theta and Psi matrices in Tech1.
Daniel posted on Friday, September 03, 2004 - 10:48 am
I should have been more clear, I mean specifications for the residual variances for the binary distal outcomes, after they have been regressed on the covariates in the ggmm. My goal is to tell whether there is a statistically significant difference in the binary distal outcome for smoking in 12th grade based on class. I regressed the binary distal outcome on prior smoking (baseline) in the overall model.
bmuthen posted on Saturday, September 04, 2004 - 5:03 pm
Binary outcomes have no variance parameters - compare with logistic regression where you only estimate an intercept and a slopes, no residual variances. Hope this answers your question.