I might be missing something obvious, but as far as I can tell, this isn't explicated anywhere. How does OpenMx deal with missing data for WLS? Is it listwise/pairwise or is the polychoric covariance matrix computed using FIML?
I might be missing something obvious, but as far as I can tell, this isn't explicated anywhere. How does OpenMx deal with missing data for WLS? Is it listwise/pairwise or is the polychoric covariance matrix computed using FIML?
mxData offers an naAction option. If you don't exclude missing data then it is ignored pairwise. There isn't really a correct way, math-wise, to handle inconsistent sample sizes in a covariance matrix. It just does the best it can.
HI
It's FIML-pairwise. FIML is used to estimate the polychorics, but the variables go into the ark two-by-two, so it isn't FIML. Note, however, that this is likely a lot better than listwise deletion throughout.