Question about RMSEA and multigroup model comparison

[Kam Tai Tony Kwan]

Hello all,

Firstly, I am having a question about RMSEA in multi-group model. I wonder if RMSEA provided in multi-group model has already been corrected w.r.t. no. of groups, suggested by Steiger (1998). If not, it would be grateful to describe the formula used for RMSEA in lavaan/ any procedure to further correct that.

Another question is that if I want to conduct chi-sq diff. tests between single group vs multi-group model, I wonder if this is a good procedure using lavaan:

1. multi-group model, but setting constraint group.equal = c("intercepts", "regressions", "residuals", "residual.covariances") as a single group model

2. multi-group model, freeing the constraints as multi-group model

3. conduct anova(model_1, model_2)

This arose because I cannot conduct anova between a simple single group model and a multi-group model (as it said the meastructures were different)

And also, I don't know why the multi-group models produced non-positive definite vcov matrix using MLR while there was not problem using ML. It would be great if any advice could be provided

Sorry for the long questions and thank you! I am looking forward to your advice.

[car...@web.de]

The answer to your first question can be found in the source code:

#  multiple group correction
#  note: recent builds of EQS also use this 'correction'
#        perhaps we should have an option to obtain the 'old' one
#if(object@Options$mimic %in% c("Mplus", "lavaan")) {
   RMSEA <- RMSEA * sqrt(G)
   if(scaled) {
      RMSEA.scaled <- RMSEA.scaled * sqrt(G)
      RMSEA.robust <- RMSEA.robust * sqrt(G)
   }
#}

In short: Yes, the correction will be made. Regarding your second question. This is the usual procedure. Maybe you should consider the permutation test?

Gesendet: Donnerstag, 06. Januar 2022 um 09:49 Uhr
Von: "Kam Tai Tony Kwan" <kwanka...@gmail.com>
An: "lavaan" <lav...@googlegroups.com>
Betreff: Question about RMSEA and multigroup model comparison

--
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 view this discussion on the web visit https://groups.google.com/d/msgid/lavaan/f01f7c8f-8208-4fb7-a503-a152ffff0ee8n%40googlegroups.com.

 

 

[Kam Tai Tony Kwan]

Thank you for your advice and I find it very useful!

car...@web.de 在 2022年1月6日 星期四下午7:59:40 [UTC+8] 的信中寫道：