You are here

goodness-of-fit indices for ML SEM

3 posts / 0 new
Last post
jkraemer's picture
Joined: 12/18/2022 - 02:50
goodness-of-fit indices for ML SEM

I have been asked by a reviewer to offer a better justification for the fit of my proposed ML model by "indicating the fit of the within and between model separately". After reading several chapters made for MPlus, I first looked to mxRefModels(). But, I am under the impression that it does NOT work with ML SEM. So, I tied to make the saturated and independence models for within and between myself for the various calculations. I was successful regarding the independence model but not the saturated model.

Through different combinations of the within and between parts of the independence, saturated and target/candidate model, I have noticed that all the attempts from mxTryHard() leads to "All fit attempts resulted in errors - check starting values or model specification" when I include the saturated within part with any of the above between parts. So, I am convinced the error is connected to the within part... either my code or something else.

I would be happy to include my code for your review... but basically I am trying to do what Rappaport colleagues ( have claimed to have already done in a 2019 paper. This is documented in their Supplementary OpenMx RCode: [MSEMfitFunctionAppendix.R]. There are a few minor easy-to-spot syntax errors in the code as well as some (what I think are) very clear missteps in referencing (?current?) OpenMx Objects. With no other options, I will continue to try and rewrite a bunch of this RCode to "make it work". But, does anyone have a functioning version of this code or a substitute? How else might I conduct fit tests of the within and between part separately?

Thx in advance...

AdminNeale's picture
Joined: 03/01/2013 - 14:09
I'll ask Lance...

To see if there's a functioning version of the code available.

Lrappaport87's picture
Joined: 11/28/2015 - 15:44
Code Updated

Thank you for spotting that the original code no longer worked with updates to OpenMx or R. Thank you for also catching the minor syntax error, an accidentally deleted quotation mark in the code. I updated the code so that it now references the fitted model in a different way when computing model fit indices. This approach should work with current implementations of OpenMx and R. That said, I am limited to testing it on a MacBook Pro using a M1 Pro chip with R version 4.1.2 and OpenMx version 2.20.3. Please let me know how it goes with the updated code and feel free to reach out if you run into any additional challenges (