I trying to estimate a model with ML a categorical outcome by several predictors (dummies and continuous). I have no missing values on the outcome variable, but missings on the predictors. I'm using Mplus version 7, so I am expecting FIML by default. However, Mplus keeps on excluding half of my cases and claims they are missing on the DV.
Missing data theory applies only to dependent variables not to covariates. If you mention the variances of the covariates in the MODEL command, they will be treated as dependent variables, that is, distributional assumptions will be made about them.
Jas Wer posted on Sunday, March 08, 2015 - 12:17 pm
I'm running a bivariate Cholesky twin model, using the script provided on the Mplus website (http://www.statmodel.com/examples/genetics.shtml). For no apparent reason, Mplus keeps excluding cases from my analysis and I'm not sure what to do about it. My script is the same as this:
[height1 height2] (mh); [weight1 weight2] (mw); height1 height2 (v11); weight1 weight2 (v22); height1 with weight1 (v12); height2 with weight2 (v12);
height1 with height2 (m11); height1 with weight2 (m12); weight1 with height2 (m12); weight1 with weight2 (m22);
model DZ: height1 with height2 (d11); height1 with weight2 (d12); weight1 with height2 (d12); weight1 with weight2 (d22);
I'm using the MLR estimator (although the problem also happens if I don't use it). Any advice would be greatly appreciated.