Jon Heron posted on Sunday, February 07, 2016 - 12:53 am
I have some daily self-report data with a strong autocorrelation. Unsurprisingly, a simple longitudinal mixture model is unable to properly explain this (as is evident from Tech10) so I thought I would try out the relatively new "rescov" setting.
The first thing I tried was to change to this parameterization but not introduce any additional parameters - to check I was obtaining the same solution as I had previously. I discovered this was not the case. Even if I use the results from my standard model as starting values I cannot get near the same solution.
I will keep plugging away at this, but was wondering if you had any tips?