(Residual) Dynamic SEM (RDSEM/DSEM) in lavaan

[MH Manuel Haqiqatkhah]

Hi,

I am working with (highly) autocorrelated hierarchical time series. I need to fit a (residual) dynamic SEM (RDSEM/DSEM) model to the data. In order to make the matter clear, I give an introduction to DSEM and RDSEM.

Intro to (R)DSEM:

In short, DSEM is suitable for intensive longitudinal data (ILD) where the dynamics of (latent) variables of an ILD (e.g., +50 measurements per individual) and their autoregressive property is modeled. There is quite a lot of literature on DSEM using Mplus (e.g. Hamaker et al, 2018 and McNeish, 2018). However these two do not include a measurement model in their DSEM.

However, sometimes there still remains an autoregressive effect among the residuals. It is basically the idea behind the dynamic factor models (DFMs) that has been discussed in that literature (cf. Poncela & Ruiz, 2012). In the context of SEM, this has been briefly discussed in (Asparouhov et al. 2018) and in more details in (Asparouhov & Muthén, 2018). DSEM and RDSEM are only implemented in Mplus 8.1.

To the best of my knowledge, the only attempt to model DSEM in R is ctsem package by (Driver, Oud, & Voelkle, 2017) where a measurement model is included but it overlooks residual AR. More theoretical details can be found in Driver's (2018) dissertation.

What I'm doing it at the moment:

I want to implement (R)DSEM models in r using lavaan. I am fitting a model to lagged pairs of measurements as I explained in the beginning of this question, and "manually" (using what was explained here in my other question) including the autoregression among factor and among residuals in the lavaan model definition. However, it is very cumbersome and time-consuming, and the model checking and alternative model comparisons requires quite a lot of effort. 

What I'm looking for:

Is there an easy/more efficient way of doing this in lavaan (or elsewhere)?

Thanks in advance,

M.H. Manuel.

[Edward Rigdon]

Stephen Boker at Virginia shared applications of OpenMx that make it easier to estimate DSEM models. He did a one-day at Modern Modeling Methods conference at UConn a few years back. His page has some links:

http://people.virginia.edu/~smb3u/  

Given these special functiuons, it is easy to implement the rest via lavaan, if you really don't have a taste for OpenMx's detailed syntax. I can share some scripts, if you are interested.

--Ed Rigdon

--
You received this message because you are subscribed to the Google Groups "lavaan" group.
To unsubscribe from this group and stop receiving emails from it, send an email to lavaan+un...@googlegroups.com.
To post to this group, send email to lav...@googlegroups.com.
Visit this group at https://groups.google.com/group/lavaan.
For more options, visit https://groups.google.com/d/optout.

[MH Manuel Haqiqatkhah]

Thanks for your reply.

I looked at Boker's slides and materials in that conference and they are fantastic! However, as far as I could find, there are not much measurement models implemented in the examples (let alone AR of residuals.)

I would appreciate if you can send me the OpenMX scripts.

Thanks,

Manuel.

To unsubscribe from this group and stop receiving emails from it, send an email to lav...@googlegroups.com.