Procedure to limit problems of local ... PreviousNext
Mplus Discussion > Growth Modeling of Longitudinal Data >
 Rowan Smeets posted on Friday, June 22, 2018 - 3:00 am

I am a PhD student currently running a Latent Class Growth Analysis (liniear one) and I would like to check with you whether I follow the correct procedure for limiting problems of local solutions.

Next I will explain the procedure I follow:
First I run the LCGA with random starts of 20 4 (without tech11 to limit computation time). I check whether the output tells me that the best loglikelihood has been replicated. If it says so, I rerun the input with a starts value of 100 20 in order to be sure that the best loglikelihood is again replicated. If so, I put my random starts at 0, add "optseed" = the seed value from the best loglikelihood found with starts 20 4, to the ANALYSE part in the input and add tech 11. So I always take the seed value from the first run (with a specific value of starts) in which I replicate the best loglikelihood. This seed value then belongs to the first loglikelihood in the first run in which is succesfully replicate loglikelihood.

I am happy to hear whether this is the optimal procedure.

Thank you in advance.
 Bengt O. Muthen posted on Friday, June 22, 2018 - 3:53 pm
See our Web Note 14.

Starts = 20 4 is often too little.
Back to top
Add Your Message Here
Username: Posting Information:
This is a private posting area. Only registered users and moderators may post messages here.
Options: Enable HTML code in message
Automatically activate URLs in message