Different results from different vers... PreviousNext
Mplus Discussion > Growth Modeling of Longitudinal Data >
Message/Author
 Seung Bin Cho posted on Thursday, September 04, 2014 - 8:30 am
Hello, Dr. Muthen.
I have different results from two different versions (v5 and v7.11) of Mplus running an exactly same script with the same data.

From version 5 with 2-class solution, maximum likelihood was replicated and two classes with proportions of .16 and .83 were estimated.

From version 7.11 with 2-class solution, maximum likelihood was not replicated and, in the final solution, two classes with proportions of <.001 and .999 were estimated.

Interestingly, in the 3-class solution using v7.11, maximum likelihood was still not replicated, and three classes with proportions of <.001, .16 and .83 were estimated. Means of I and S factors in the classes with proportions of .16 and .83 were almost identical with that from 2-class solution from version 5.

I think such almost non-existing classes from v7 are some kind of artifact but am not sure what caused them.
I also wonder what results are more trustworthy.
If that almost non-existing class from v7 is an artifact, how would I check if it's an artifact, and what advice would you recommend to present the result?

Thank you very much for your help.
 Bengt O. Muthen posted on Thursday, September 04, 2014 - 3:50 pm
You should not interpret results unless the best logL is replicated several times.

To ensure the best performance, you should always use the latest Mplus version which is now 7.2.
Back to top
Add Your Message Here
Post:
Username: Posting Information:
This is a private posting area. Only registered users and moderators may post messages here.
Password:
Options: Enable HTML code in message
Automatically activate URLs in message
Action: