I am working on a multilevel mediation model and I get this error essage:
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 0.533D-27. PROBLEM INVOLVING THE FOLLOWING PARAMETER: Parameter 1, %WITHIN%: [ PARMASCL ]
THE NONIDENTIFICATION IS MOST LIKELY DUE TO HAVING MORE PARAMETERS THAN THE NUMBER OF CLUSTERS. REDUCE THE NUMBER OF PARAMETERS.
PARMASCL is the mediator in my model. Because it is group-mean centered, it has a mean of zero. When I fix its intercept to zero, the error message disappears.
I have two questions about this: 1. Is this a correct solution? 2. What is causing the error message?
Is it because I am estimating the intercept of a group mean centered variable, or does this have to do with my particular model and data?
(Intuitively it makes sense that estimating an intercept that is known to be zero causes a problem, but what is the reason? Could you maybe point me in the direction of an answer to that question?)
Hopefully you can help me understand this, Gerald.