any prediction option for lavaan.mi with categorical data?

[Michael]

Dear colleagues,

thanks for your work and support (already in advance).

I'm faced with some rather messy data (especially quite some missings), consisting of five ternary indicators, one factor and three groups. 

So far, testing for invariance worked fine with imputed data (using amelia). 

The next step actually should have been (following e.g. Leite, 2017) to use the predict-function to get some factor scores for propensity score analysis. 

(I know using the factor scores might not be the best approach, but the psm is not really negotiable)

Problem is: predict seems not to like lavaan.mi-objects, plausibleValues does not like categorical data...and besides the fact, that i'm not familiar with the sam-approach (which sounds like a possibility) it seems also not to take mi-objects (yet)

So, as i am stuck and feel out of options: Did I miss another option for getting scores with multiply imputated categorical data? Or is there any other path to follow? 

Thanks a lot,

Michael

[Michael]

...even a qualified "no" would actually help...in some way...

[Terrence Jorgensen]

...even a qualified "no" would actually help...in some way...

You have correctly identified the limits of the software.  lavaan has a predict() method for lavaan objects, but semTools does not have a predict() method for lavaan.mi objects.  Because of the inherent limitations of using factor-score estimates as known values in subsequent analyses, I have not been motivated to develop a predict() method for lavaan.mi objects, but I did provide plausibleValues().  However, lavaan does not provide the SEs or asymptotic covariance matrix of factor score estimates for the methods available for categorical indicators; thus, there is no mechanism for plausibleValues() to do its job in that context.  

Mplus would be able to provide plausible values for categorical indicators themselves (LRESPONSES), as well as for factor scores, using Bayesian estimation with categorical indicators:  https://www.statmodel.com/download/Plausible.pdf  

Terrence D. Jorgensen

Assistant Professor, Methods and Statistics

Research Institute for Child Development and Education, the University of Amsterdam

http://www.uva.nl/profile/t.d.jorgensen