Multigroup multilevel SEM - is this possible?

[Robert Udale]

Hello,

I've been asked to implement the following SEM model:

social_id_model = '

 level: 1

        

        # Latent variables (=~)

        injunctive_norm_general =~    1 * Q36_1 + 1 * Q37_1

        descriptive_norm_general =~   1 * Q38_1 + 1 * Q39_1

        personal_norm_general =~      1 * Q53_1 + 1 * Q54_1

        policy_acceptance         =~  1 * Q64_1 + 1 * Q66_1

        # Regressions (~)

        policy_acceptance ~ Q63_1                

        Q63_1 ~ injunctive_norm_general + descriptive_norm_general + personal_norm_general

        personal_norm_general ~ injunctive_norm_general + descriptive_norm_general

        injunctive_norm_general ~ Q35_1

        descriptive_norm_general ~ Q35_1

        level: 2

        

        # Latent variable

        injunctive_norm_general =~    1 * Q36_1 + 1 * Q37_1

        descriptive_norm_general =~   1 * Q38_1 + 1 * Q39_1

        personal_norm_general =~      1 * Q53_1 + 1 * Q54_1

        policy_acceptance         =~  1 * Q64_1 + 1 * Q66_1

        

        # Regression.

        Q63_1 ~ injunctive_norm_general

        

        # Residual correlations

        injunctive_norm_general ~~ descriptive_norm_general + personal_norm_general

        descriptive_norm_general ~~ personal_norm_general

      

'

social_id_fit = sem(social_id_model, data,  cluster = "country_sample", group = "gWP4")

The data is comprised of responses from 31 countries (country_sample) and is split into three groups (gWP4). 

When I run this code, we get the following error message:

Error in ov.names.l[[g]] : subscript out of bounds
 

What could be causing this error?

Is it possible to do a multi-level and multi-group analysis like this?

[jpma...@gmail.com]

Hello,

Why the 1*item in your mearusment models? Try to run the models without the “1”…

Best.

Jº

--
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.
To view this discussion on the web visit https://groups.google.com/d/msgid/lavaan/55d4a677-b68b-45f0-a784-f67fa64903b7%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Terrence Jorgensen]

Is it possible to do a multi-level and multi-group analysis like this?

Undocumented so far, but yes.  Groups must be considered as fixed, and vary only at Level 2.  So you need to specify blocks for each group, and within each group's block, you specify the Level 1 and Level 2 blocks.  

model <- '
Group: 1

Level: 1

  foo ~ bar

Level: 2

  foo ~ bar



Group: 2

Level: 1

  foo ~ bar

Level: 2

  foo ~ bar


 
Group: 3

Level: 1

  foo ~ bar

Level: 2

  foo ~ bar
'

But you already have a small Level 2 N=31, so splitting that into 3 groups spells disaster.  Even if your model converges, you couldn't trust your results.

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

[Franzi K.]

Dear Terrence,

I was wondering which optimizers and estimators might be suitable for having a multi-group multilevel SEM? Moreover, are there any other parameters that enhance running such a model?

Thank you in advance!

Best,

Franzi

[Terrence Jorgensen]

which optimizers and estimators might be suitable for having a multi-group multilevel SEM?

Same options as for single-group MLSEM: maximum likelihood or EM algorithm.

http://lavaan.ugent.be/tutorial/multilevel.html

 

Moreover, are there any other parameters that enhance running such a model?

I don't understand your question.

[Franzi K.]

Sorry for the unspecific question, I was rather referring to things I have to specify in the sem()-function but in the end I came to the conclusion that my sample is too small anyway :-(.

Thank you!