Modification indices after Satorra-Bentler-adjusted confirmatory factor analysis?

[Jenny Rickardsson]

Dear fellow Stata users

I am doing a validation of a form (a ten-item form with seven Likert-like answering options on each item) in a new language. I have performed a confirmatory factor analysis with the Satorra-Bentler adjusted method, since my data has problems with univariate and multivariate normality. Model fit is not so good. I try to follow a process previously proposed by the original writers of this form (see reference below), and they have done Satorra-Bentler, then model modification indices and subsequent changes in the model for better fit. This was done i MPlus software. But in Stata I cannot seem to do a modification indices analysis after Satorra-Bentler this is what I get:

. estat mindices
modification indices not allowed with vce(sbentler) results



Does anyone know if it is possible in Stata to get theoretically sound modification indices with another command/analysis?

I work in Stata 15.1, on a Mac. My dataset consists of 252 observations.

Hope to hear from you!

All the best,

Jenny


Smout, M., Davies, M., Burns, N., & Christie, A. (2014). Development of the Valuing Questionnaire (VQ). Journal of Contextual Behavioral Science, 3(3), 164-172. doi:10.1016/j.jcbs.2014.06.001

[Phil Bromiley]

You didn't get a quick answer. You;ll increase your chances of a useful answer by following the FAQ on asking questions - provide Stata code in code delimiters, readable Stata output, and sample data using dataex. Remember, for the most part, we know statistics and Stata but not your substantive area so questions that are really about the substantive area are likely to not get helpful responses.

I'm not an sem person but let me try to offer some suggestions. You could use a conventional sem where Stata does provide all the modification indices. Alternatively, Estat dof will give you Satorra-Bentler chi squares - I suspect you can use this with multiple models to identify modifications. After all, what modification indices are often doing is looking for modifications that increase fist. I assume you've googled to see if such a routine exists - I couldn't find one easily. Finally, you can probably program it in Stata.

[Richard Williams]

Modification indices are basically a diagnostic technique. I don’t think diagnostics have to be accurate to 12 decimal places. Just run the model with an option that gives you MIs. Use your own judgment on making the changes — some modifications will be more sensible than others.

[Hoai My]

Hi! Did you find out how to solve this problem? I am currently having the same issue on my CFA.