Anonymous posted on Wednesday, June 29, 2005 - 8:12 am
I recently came across an (unpublished) manuscript suggesting that GMM cannot appropriately deal with freely estimated factor loadings of the growth factor, if they are allowed to vary across classes. Is that true?
In my case: I have 6 timepoints & 2 classes, with the first 3 loadings fixed to 0,1,2 while the last 3 loadings are freely estimated and allowed to vary across the two classes. In addition I have 2 time-invariant covariates assuming class varying effects on i and s. The results are as expected - is there a reason to be concerned? THANK YOU
BMuthen posted on Saturday, July 02, 2005 - 5:56 pm
This is not true. Factor loadings can vary across classes.