Justin Mark posted on Monday, September 19, 2016 - 12:40 pm
I am simulating categorical items from an underlying 2-factor structure, and then imposing a high degree of missingness in most of the items. There are 14 items in all, and one item corresponding to each factor is observed for all N. The other items are observed in only half of the sample. Imagine 2 groups that filled out 2 different questionnaires corresponding to the same two underlying factors, with two overlapping items in the surveys.
I am generating the data separately and then running the external Monte Carlo feature. When I do this on bash, on get the following error message:
forrtl: severe (153): allocatable array or pointer is not allocated
The output file contains no information or error message, but simply stops after the title.
I ran the same toy simulation script without imposing the missing data on the data generation side of things, and the script works. I have also run the same type of script on continuous indicators, no problem.
Currently I have only used the WLSMV estimator. Is there a way to fit this model with categorical outcomes and this much missing data? Would it be more effective to specify as a multi-group model?