Troubleshooting
Problem
I'm running a mixed model in SPSS MIXED, and am receiving the following warning: "The final Hessian matrix is not positive definite although all convergence criteria are satisfied. The MIXED procedure continues despite this warning. Validity of subsequent results cannot be ascertained." Can you give me any advice on what might lead to this warning?
Resolving The Problem
A common cause of this warning is a model specification that involves redundant covariance parameters. You may need a simpler covariance structure specification in order to avoid this problem. Also, failure to specify a SUBJECT variable on the RANDOM subcommand can produce a redundant covariance parameter.
In other cases, increasing the number of step-halvings allowed or the number of Fisher scoring steps may help to achieve convergence on all criteria, including the Hessian matrix. Settings for these criteria can be changed in the Estimation dialog box or via command syntax using the CRITERIA subcommand.
Failure to converge can also be a signal of redundant covariance parameters.
For a general strategy on choosing mean structures and covariance structures for mixed models, see Technote 1476569. For information on the relationships among structures and which are nested within others, see Technote 1480315.
Related Information
Historical Number
52999
Was this topic helpful?
Document Information
Modified date:
16 April 2020
UID
swg21480810