Dear Dr. Muthen, I am curious about a method I read about recently. In a study, Y individuals were followed up over time. Of these Y individuals, X died and neuropathology data was collected after their death (Y=6X, to give some idea of sample sizes). An outcome variable that had been measured during the study in the full sample of Y individuals was modeled using a growth model with factor loadings fixed at 0 and 1. The intercept and slope of the growth model were regressed on the neuropath data collected in the subsample of X deceased individuals.
Leaving aside potential issues with missing data, I am uncomfortable with the idea of regressing the intercept and slope on variables measured after the change (estimated by the slope) already happened. Similarly, with regressing the intercept.