Muthén's robust weighted least squares factoring method for binary items…in R?

I am working on an exploratory factor analysis of dichotomous items. I've found this post extremely helpful. However, Mplus seems to be a standard that journals in my field are expecting, and I would love to replicate Mplus results in R.

I've gotten my results to come very close to Mplus, using the fa function on a polychoric correlation matrix (created with the hetcor function) and the geominQ rotation. It seems that the major difference between the fa function and Mplus is that the latter uses a robust weighted least squares factoring method (WLSMV - a diagonal weight matrix), whereas the former uses a regular weighted least squares (WLS) factoring method.

Has anyone managed to use R to replicate Mplus factor analysis for binary items? Has anyone figured out how to implement WLSMV in R?

Thanks!

• aside...Mplus can also generate bootstrap standard errors - advised to try and account for the extra variability in polychorics - see ch12 of mplus examples on how to implement. – user20650 Apr 3 '13 at 19:48

1 Answer

What you want is in the lavaan package, the function is named sem. Try writing an argument estimator = "WLSMV". For more information read this.

• i dont think there is an example of efa in lavaan package but look at efa in a cfa framework example at amarkos.gr/research/lavaan for how to implement. Wish i had noticed this before. – user20650 Apr 3 '13 at 20:01
• Thanks AAA and user20650, this is looking promising -- I'm still messing around with the options for the sem function and will update here once I figure it out. – sprins123 Apr 7 '13 at 16:34