|
|
Warning about MLR standard errors |
|
Message/Author |
|
|
I am trying to run a two-level regression analysis for a continuous dependent variable with a random slope. The model is running, but I get this warning: THE MLR STANDARD ERRORS COULD NOT BE COMPUTED. THE MLF STANDARD ERRORS WERE COMPUTED INSTEAD. THIS MAY BE DUE TO NEAR SINGULARITY OF THE RANDOM EFFECT VARIANCE/COVARIANCE OR INCOMPLETE CONVERGENCE. Can you tell me what this warning means and if I should be worried about it? Is there anything I can do to solve this? Thank you very much for your help! |
|
|
First, I would make sure I am running Version 7. Second, check if all random effects really have variance larger than zero. |
|
Katrina Lin posted on Tuesday, April 15, 2014 - 6:57 pm
|
|
|
Dear Prof. Muthen, I was testing a moderating effect with IV, moderator and DV are all Level 1 variables. I used random intercepts and random slopes to test. Although I can get the estimates and significance test, the standard errors are quite different from what I got from HLM. I think it is the reason below that caused the difference. "WARNING: THE MLR STANDARD ERRORS COULD NOT BE COMPUTED. THE MLF STANDARD ERRORS WERE COMPUTED INSTEAD. THE MLR CONDITION NUMBER IS -0.600D-02. PROBLEM INVOLVING PARAMETER 9. THIS MAY BE DUE TO NEAR SINGULARITY OF THE RANDOM EFFECT VARIANCE/COVARIANCE OR INCOMPLETE CONVERGENCE." Can you suggest? I am using Mplus 6. Does version 7 help with this issue? Thank you very much! |
|
|
HLM probably uses ML - try ML. It is always a good idea to use the latest version of Mplus. |
|
Back to top |
|
|