Is it possible with OpenMx to estimate individuals' values or scores on a latent variable? For example, using the one factor model (https://openmx.ssri.psu.edu/svn/trunk/demo/OneFactorModel_PathRaw.R), is it possible to estimate what each case's value is on the latent factor? If so, how can this be done?

Thanks in advance!

Factor score estimation is a controversial topic. If you want to use a traditional (i.e., Bartlett) estimator, you could use the OpenMx estimated loadings and use the functions in the psych library.

However, this method won't work terribly well if you have missing or ordinal data. In an upcoming paper, I and Mike Neale outline a method for estimating factor scores using a model for each row of data. We hope to put together the functions for public use shortly, but I've attached the appendix to the paper that outlines the coding. The full paper should be in the next issue of MBR.

Very interesting, Ryne. Thanks for sharing. I will look forward to reading the paper when it is published. If I were to take the factor scores from the Bartlett estimator, how would I rescale them so that each person's score is on the same metric as the latent factor? The variables composing my latent factor range from 0-200, but the individuals' factor scores from the Bartlett estimator range from -2 to +2. How can I convert these factor scores into meaningful values in terms of the latent metric?

Many thanks!