Hello,

I'm interested in estimating a model in which a little over half of my 40 indicators are categorical (which are split fairly evenly between ordinal and dichotomous outcomes... I have approx. 43,000 observations and am interested in estimated 11 latent variables...). From what I've read, I should be using Yuan and Bentler's (2000) approach, which relies on scaled-chi square and robust standard errors and that this approach could be implemented in mx by specifying "estimator = MLR". I was wondering if there is an openMx equivalent for this command, or, if not, if anyone had written an openMx script implementing the approach which they would be willing to share. As I still quite quite new to CFA / SEM, I would also be curious to hear any thoughts as to whether this is still considered the best way to handle non-normal missing data.

Many thanks in advance,

Pete

Yuan K. H. and P.M. Bentler. “Three Liklihood-Based Methods for Mean and Covariance Structure Analysis with Non-Normal Missing Data.” Sociological Methodology 2000:165-200.