Jon Heron posted on Wednesday, October 07, 2009 - 7:23 am
I've been throwing a variety of models at 5 repeat measures of emotional symptoms in the hope of plucking out a bunch of interesting kids who increase from childhood into adolescence.
Today I've been playing with GMM's on a standardized continuous form of my scales.
As I stare blankly at the results, I wondering if it really makes sense to do this. I've heard of doing growth modelling on standardized data but not necessarily mixture modelling. It seems that I can't get my desired increasing class without a corresponding decreasing one to keep the population mean at zero.
If you're wondering why the scales are standardized, I've chucked my dichotomised sum-score measures for now and am experimenting with some factors derived from the individual items. I don't this this is relevant to my question though.