Factor loadings, residual variances and means estimation
Posted on
krzysiek
Joined: 09/05/2013
Dear users,
I was trying to model latent variable via the common factor model using path-centric model specification as it is shawn in the examples in OpenMx User Guide, Release 1.2.0-1919, chapter 2.2 Factor Analysis, Path Specification (http://openmx.psyc.virginia.edu/docs/OpenMx/latest/OpenMxUserGuide.pdf). My purpose was to to estimate factor loadings, residual variances and means. But in all my attmpts I did not receive the chi-square statistics and some other goodness of fit measures like CFI, TLI and RMSEA. Even in the example exactly copied from the above source (http://openmx.psyc.virginia.edu/svn/tags/stable-1.2/demo/OneFactorModel_PathRaw.R) I have not received the chi-square value. Is there any mistakes?
best regards,
Krzysiek
If you were using raw data,
If you were using raw data, that is mxData with type="raw", then these fit indices are not returned by default so that the model estimates faster. These fit indices require a best-fitting comparison model that we call the Saturated Model. By default, OpenMx does not fit this model for you because it often doubles the amount of run-time. The saturated model is a model with all the means and covariances freely estimated as parameters. Future versions of OpenMx will have a helper function for making this model. I've attached the R file that defines this function for you to use. Running the attached R file defines a function called mxSaturatedModel that takes a simple mxModel, builds, and optionally fits the corresponding saturated model.
The general format of use is as follows:
# Example usage
amod <- mxModel(blah)
smod <- mxSaturatedModel(amod, run=TRUE)
summary(amod, SaturatedLikelihood=smod)
Once the SaturatedLikelihood argument is given the saturated model, the summary should return all the fit indices you mentioned. Let us know if it does not!
Log in or register to post comments
In reply to If you were using raw data, by mhunter
Cool
Mike H, that's very nice! I hope a multiple group version will be forthcoming :).
One thing I noted, your file defines omxSaturatedModel() not mxSaturatedModel() so a slight tweak to read
# Example usage
amod <- mxModel(blah)
smod <- omxSaturatedModel(amod, run=TRUE)
summary(amod, SaturatedLikelihood=smod)
seems needed.
Log in or register to post comments
In reply to If you were using raw data, by mhunter
Thank you!
Thank you very much!
It really helps me!
Krzysiek
Log in or register to post comments