Entering the number of observed parameters

Posted on Mon, 05/30/2011 - 20:31

Sacha Joined: 05/30/2011

Attachment

Size

MxOrdModel.R

3.48 KB

Forums

Confirmatory Factor Analysis and Measurement Models

Hi,

This is literally the first day I every used OpenMx so it is all quite new to me, maybe I am making a stupid mistake here.

I am attempting to perform a two-group CFA of one continuous factor on six ordinal predictors in OpenMx using Weighted Least Squares (WLS) estimation. I have the polychoric correlation/covariance matrices and I managed to get it working with help of:

http://openmx.psyc.virginia.edu/docs/OpenMx/latest/FactorAnalysisOrdinal_Matrix.html
http://openmx.psyc.virginia.edu/thread/530

It runs and I get parameter estimates that look reasonable (pretty much the same I get from mplus, so I am happy with that). Not sure if the weights matrix is estimated correctly (I now do adf.wmat() on the raw data) but that's not that much of a problem right now.

Only thing that I don't get working is the goodness of fit indices. I can't figure out how I tell the fit how much observed parameters and observations I have. Is this possible? Or should I make changes to how I setup the model?

Attached is my model. Thanks in advance!

Replied on Mon, 05/30/2011 - 21:41

mspiegel Joined: 07/31/2009

Hi. Welcome to OpenMx! I

Hi. Welcome to OpenMx! I think you're looking for the 'numObs' or 'numStats' arguments to the summary function. See ?summary for more information.

Log in or register to post comments

Replied on Tue, 05/31/2011 - 02:28

Sacha Joined: 05/30/2011

Thanks! That works, I only

Thanks!

That works, I only still get no estimates for the goodness of fit indices (like chi2 and RMSEA), but that makes sense since I don't say what my observed and expected values are. I can just compute those manually.