| Message/Author |
|
|
|
|
Dear Drs. Muthen and Muthen,
I am working on a LCGA in which i have count variables measuring the number of months spent in unemployment at each time point from age 15-31. The frequency table is too large so I can't look at bivariate residuals.
Event though conditional independence is assumed at the level of class k, would it be possible to model an autoregressive structure within the LCGA framework? Or are there any other methods to address local dependence in LCGA?
Thank you in advance |
|
|
|
|
Growth mixture modeling allows deviations form local independence - see UG examples.
Autoregressive structure gives complex computations with counts. You can introduce a factor for each adjacent pairs of outcomes but that leads to high-dimensional numerical integration with ML. |
|
|
|
|
| Thanks a lot for the advice. Would it be possible to specify dependence with just one factor measured by all outcomes? |
|
|
|
|
| Yes, that would be a more restrictive auto-correlation structure, but ok. |
|
|
| Back to top |
