Random-Intercept Cross-Lagged Panel Model with auxiliary variables

[Ulrich Schroeders]

Hi!

I'm estimating a random-intercept cross-lagged panel model (RI-CLPM; Hamaker et al., 2015) and need to incorporate some background variables (age, sex, education) into the imputation model to improve fiml-based handling of missingness. Using the auxiliary function of the semTools automatically creates correlations between the observed variables and the covariates (covariates ~~ obs_vars).

But since the latent variables are single indicator variables, I wonder if the correct way to specify this model, is to correlate the covariates to all latent variables in the model (covariates ~~ lat_vars). This seems to result in reasonable parameter estimates and only slight corrections as I would have been expected because I'm improving the imputation model only.

Thanks in advance for any comment, Kind regards

Ulrich

[Ulrich Schroeders]

I have to add/correct two things:

First, I should have written "analysis model" instead of "imputation model", since we are talking about FIML.

Second, the X^2 is identical, also parameter estimates for all effects. What differs are df (and fit statistics) and, of course, the relations to the covariates.

Kind regards

Ulrich

[Atanas Janackovski]

Have you seen this: https://jeroendmulder.github.io/RI-CLPM/lavaan.html

This covers covariates... Not sure if it applies the same restrictions as what you're requiring/suggesting

[Ulrich Schroeders]

This is a great resource, I'm familiar with.

I think the question about the auxiliary-variables-fiml-thing isn't cover here.

Nonetheless, thx.

Kind regards

Ulrich

[Atanas Janackovski]

No worries. I'm interested in learning more  about your approach. Can you point me in the direction of something that might help me understand and implement it?

[Atanas Janackovski]

Found it... https://www.rdocumentation.org/packages/semTools/versions/0.5-4/topics/auxiliary

[Atanas Janackovski]

Hey Ulrich,

I am now trying something similar to what you are suggesting in this post. I'm wondering whether you freely estimated the means for your covariates in your syntax, i.e., 'covariates ~ 1'?

Cheers,

Atanas