Hi, I used LCA (on 21 items) to estimate a one class, a two class, and a three class model with Bootstraps (BLRT), these all worked. Then, I estimated a four class model with Bootstraps (ANALYSIS: TYPE = MIXTURE; starts = 500 5; OUTPUT: TECH1 TECH8 TECH14;) and I got the following warning:
WARNING: THE BEST LOGLIKELIHOOD VALUE WAS NOT REPLICATED IN 3 OUT OF 5 BOOTSTRAP DRAWS. THE P-VALUE MAY NOT BE TRUSTWORTHY DUE TO LOCAL MAXIMA. INCREASE THE NUMBER OF RANDOM STARTS USING THE LRTSTARTS OPTION.
I have tried various different starting values (50, 100, and 500) but keep getting this message indicating that changing the starting values did not change anything. I wonder how I can solve this warning, do you have suggestions?
Jon Heron posted on Wednesday, July 14, 2010 - 1:41 pm
are you changing the right thing here?
The warning mentions LRTSTARTS but your snippet of syntax only mentions STARTS which is something else.
I get this warning often, but find it goes away if I fiddle with LRTSTARTS and LRTBOOTSTRAP e.g.
Your LCA model is rather ambitious though, even if your 21 items are binary you would need a sample size of millions to achieve anywhere near the possible number of reponse patterns. I wonder if it's worth an initial step in which you trim this list down a bit.