Message/Author |
|
Mike Barber posted on Tuesday, January 31, 2006 - 10:04 am
|
|
|
Dr. Muthen, I have fit a growth mixture model with 2 covariates and wondered how to evaluate the strength of the covariates as predictors of class membership. |
|
|
Example 8.1 shows the regression of the categorical latent variable on a covariate. You can extend that to two covariates. This is a multinomial logistic regression. If you have three classes, you need to specify two regressions. The last class is the reference class. For example: c#1 ON x; c#2 ON x; |
|
Mike Barber posted on Tuesday, January 31, 2006 - 10:55 am
|
|
|
Thanks for the reply, how can I evaluate the usefulness of the covariates? Is there something like an R-squared or strength of effect measure? |
|
|
You can look at their statistical and practical significance. As far as I know, there is no R-square avaiable for multinomial logistic regression. |
|
Back to top |