Comparing RESIDUAL and TECH10 PreviousNext
Mplus Discussion > Latent Variable Mixture Modeling >
Message/Author
 Jay Verkuilen posted on Thursday, September 29, 2016 - 4:37 pm
I've been running a lot of ordinal mixture CFA recently. The OUTPUT RESIDUAL option generates one- and two-way marginals with predicted probabilities and standardized z-scores. TECH10 also generates this but the two sets of outputs don't appear to be the same.

(1) TECH10 seems to generate marginals not stratified on the classes? Is this correct? I saw elsewhere the formula used for the standardized residuals and have been able to replicate them closely.

(2) How are the standardized residuals computed by RESIDUAL, which are broken out by latent classes? I can't seem to replicate them, even accounting for rounding error from only having three decimal places.
 Tihomir Asparouhov posted on Friday, September 30, 2016 - 9:53 am
(1) Yes

(2) sqrt(Nk)*Residual/sqrt(Estimated*(1-Estimated)), where Nk is the number of observations in class k picked up from this output section

FINAL CLASS COUNTS AND PROPORTIONS FOR THE LATENT CLASSES
BASED ON ESTIMATED POSTERIOR PROBABILITIES
Back to top
Add Your Message Here
Post:
Username: Posting Information:
This is a private posting area. Only registered users and moderators may post messages here.
Password:
Options: Enable HTML code in message
Automatically activate URLs in message
Action: