Confirmatory factor analysis

[Juan Pablo Alzate-Granados]

Hi all,

We are conducting a scale validation and we want to perform a confirmatory factor analysis
We have seen that in other publications using the LISREL 8.8 program, but we want to know if it is possible to perform this analysis using STATA with the confidence that the results will be the same

Thanks a lot

[Clyde Schechter]

Confirmatory factor analysis is just a particular type of SEM. If you look in the [SEM] manual, examples 1, 3, and 15 will give you a pretty comprehensive view of how to use the -sem- command for this.

If you are using a version of Stata older than 12 (which you should have told us in your post), then you will not have an -sem- command. But there is a user-written command -confa-, by Stas Kolenikov that will serve your needs. It is available from ssc.

[ben earnhart]

When you go to run your models (assuming Stata 12 or 13), Stata actually by default models means and intercepts, whereas Lisrel (by default) only models covariances. So they *can* get the same results, but you need to be careful about model specification. Lisrel gives you a much broader array of fit indices, but Stata gives you the ones that matter, and if you pull out a SEM textbook, you can generally compute most of the others if you want them. I'm a Lisrelite (and have been using Mplus lately), but Stata 12 and 13 make it really, really easy to model things. Maybe too easy. So before running your CFAs, do your homework so you know what you're getting into. They *can* get the same results, but you need to specify the model the same.