ACE model with repeated measurements

This is one case in which the Cholesky model is actually interpretable as a model, rather than as a convenient trick to keep estimated matrices positive definite. See http://www.vipbg.vcu.edu/~vipbg/GE/GE_OpenMx.shtml for a relatively comprehensive multivariate twin script (it fits saturated models as well). Look for the section

# Fit Multivariate ACE Model with RawData and Matrices Input
# -----------------------------------------------------------------------
multiCholACEModel <- mxModel("multiCholACE",

for the so-called 'Triple Cholesky' model. Order your variables so that time1 is variable 1, and time2 is variable2. The magnitude of the A/C/E components specific to time 2 reflects the degree of new variance at time 2 not caused by factors operating at time 1.

Thanks to Hermine Maes for providing this resource!