Cluster leads to unidentified model PreviousNext
Mplus Discussion > Multilevel Data/Complex Sample >
Message/Author
 Youyou Tao posted on Tuesday, June 06, 2017 - 3:36 pm
Hi,

I have a dataset with data from 7 states in the US. So I am using state to cluster my data (cluster = state) and used type= complex in the analysis part.

However, I always get this message: "THE STANDARD ERRORS OF THE MODEL PARAMETER ESTIMATES MAY NOT BE TRUSTWORTHY FOR SOME PARAMETERS DUE TO A NON-POSITIVE DEFINITE FIRST-ORDER DERIVATIVE PRODUCT MATRIX. THIS MAY BE DUE TO THE STARTING
VALUES BUT MAY ALSO BE AN INDICATION OF MODEL NONIDENTIFICATION. THE CONDITION NUMBER IS -0.654D-16. PROBLEM INVOLVING THE FOLLOWING PARAMETER: Parameter 7, [ B_4YI ]

THIS IS MOST LIKELY DUE TO HAVING MORE PARAMETERS THAN THE NUMBER OF CLUSTERS MINUS THE NUMBER OF STRATA WITH MORE THAN ONE CLUSTER."

So, my first question is whether I can trust my result if I get this message.

After I removed the type= complex, this error message disappeared and my model runs fine.

So, my second question is whether I should code different state into binary variables and run the state effect as co-variance instead using the cluster?

Thank you,

Yoyo
 Bengt O. Muthen posted on Tuesday, June 06, 2017 - 6:17 pm
7 is much too few clusters - you need at least 20. So, no, you cannot trust the SEs. Amd, yes, it is better to turn those tracts into dummy variables and use as covariates.
 Youyou Tao posted on Tuesday, June 06, 2017 - 6:51 pm
Thank you so much for your response. This is very helpful!

Yoyo
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: