Proc Calis error: The sample covariance or correlation matrix is not positive definite.

[gradstudent]

I am receiving an error message when trying to run a proc calis statement: The sample covariance or correlation matrix is not positive definite.

I'm unsure how to corrrect the issue.  Below is a screen shot of my code and error.  If anyone has any suggestions, I'd be willing to try them!  I'm wondering if the issue doesn't have to do with missing values in my variables.  But I'm a novice, and haven't ran a proc calis before.  Thanks in advance for any help!

[SteveDenham]

I am not a PROC CALIS user, but whenever I see that a matrix is not positive definite, two things come to mind.  You have a complete dependence between two (or more) columns in the matrix, and either the model is overspecified or there is a linear dependence in the data.  Check for collinearity in your variables using PROC REG--if everything looks OK, then it is a model specification problem.  If there are majorly collinear variables, then try using just one of them (with adjustment to your path model, as needed).

Steve Denham

[Phoenix_LJ]

Yes, you are right.