|
|
Multiple Group MIMIC model with censo... |
|
Message/Author |
|
Anonymous posted on Wednesday, March 23, 2005 - 4:22 pm
|
|
|
Hi, I'm trying to run a multiple group MIMIC model that has censored indicators for the latent factor. I have three covariates and would like to use the TYPE = MISSING command, but the program will not run and says to use TYPE = MIXTURE with KNOWNCLASS, but this causes one of the factor loadings to be negative and the program will not run when I try to model the covariates, it says that the latent factor is not a latent factor. Any suggestions? Thanks |
|
|
You will need to send your full output, data, and license number to support@statmodel.com to get an answer to this one. |
|
Herb Marsh posted on Thursday, September 25, 2008 - 5:15 am
|
|
|
I am running a MIMIC model as an alternative to a multigroup model ESEM (using data from your example 4 in your addendum for versions 5.1). TITLE:MPEX4 MIMIC 4c F1-F2 on group; DATA: FILE IS example4.dat; VARIABLE: NAMES ARE y1-y10 group; !GROUPING IS group (1 = g1 2 = g2); MODEL: f1-f2 BY y1-y10 (*1); f1-f2 on group; OUTPUT: TECH1; tech4; mod(0); sampstat; Is it possible to merely allow the grouping variable to be correlated with the EFA factors? I tried replcing the line: f1-f2 on group; with each of the following: f1-f2 with group; group with f1-f2; but got the following warnings THE STANDARD ERRORS ... AN INDICATION OF MODEL NONIDENTIFICATION. CONDITION NUMBER IS 0.896D-12. PROBLEM INVOLVING PARAMETER 43. (parameter 43 is the variance of the grouping variable) |
|
|
I think the message is because group is a binary variable and the mean and variance are not orthogonal. If you obtain standard errors, you can ignore this message. |
|
Back to top |
|
|