I calculated covariance matrix using pairwise deletion and input it into AMOS. I got one qualitative variable and missing values in the data set. I got a warning "sample moment matrix is not positive definite"
A dispersion or covariance matrix should be positive definite in order for it to have a generalized inverse. Otherwise the system in question may not be solvable. That is, the model may not be estimable. I guess that in this case the existence of missing values is the reason why you have a non-
positive definite covariance matrix. How did you handle those missing values?
Thanks for your answer Ette. I did not handle missing values. I wrote pairwise deletion when writing syntax. Isn't it a solution? I did not apply imputation because one of my variables is qualitative. And I can not use listwise deletion because sample size is small.
- Badges
- Science topic
- Similar topics
- Mathematics
- Algebra
- Matrix