Message/Author |
|
|
Hello, I am running a model and recieved messages about increasing the number of bootstrap starts. I've been increasing them with the lrtbootstrap command. I'm up to 200. At this point I'm getting non-replication of the best loglikelihood in 129/200 bootstrap draws. Is there a maximum percentage that is acceptable before I can stop adding draws, and may trust the p-value? |
|
|
Could you please send the output to support so they can diagnose? |
|
|
Hi, I am running LMR and BLRT tests using tech11 and tech14 to compare a 3-class with a 2-class mixture model. Is there any way to specify the starting values for the k-1 starts model ? I use optseed to specify my starting values for the k-class model and increased the K-1 starts for tech11 as well as the LRTstarts for tech14, however Mplus does not find my best k-1 class model that I obtained when running the k-1 class model without using tech11 or tech14. And the estimation already takes weeks... Thank you very much for your help. |
|
|
We don't have a way to enter the starting values for the k-1 class model. My advice would be this. If your 2-class model is so difficult to find you should probably not entertain the idea for a 3-class model. I would also recommend using BIC, running smaller models, running simpler models that are easier and faster to estimate. Eliminate what is not needed. |
|
Back to top |