Dear colleagues, I am having some issue with confirmatory factor analysis in AMOS. We conducted a confirmatory factor analysis on 10 factors. However, AMOS does not run the model stating that the "sample moment matrix is not positive definitive". I basically followed the James Gaskin instructions: https://www.youtube.com/watch?v=rOeFJKrGhlM
Does anybody know how to cure this issue?
Lots of thanks in advance!
As an analyst, here are the things I check when I get that error
1) Sample size - how many cases for how many items?
2) missing data - if it is extreme, even pairwise deletion will not help. You can plug with the mean of the item or else (better) use a bootstrapping method to impute missing values based on the other data you have.
3) Colinearity - Centering the items helps, but if you have strong correlations between two items (for example), create a composite of them to try in the factor analysis.
4) Non-normal distribution - sometimes you will get this error if the data are strongly positively or negatively skewed. You can use transformations to fix some skew, but sometimes you have to categorize your measure instead (and then AMOS needs a different protocol).
Good luck!!
I agree with Rimantas and with Julia. Also, check if there is reversal items and see the attach for estimation method
Dear Sudong Shang,
I agree with above suggestion. You should first check if you have a sufficient N. It might be checked by the numbers of latent variables, items, and so on...
Probably you can impose a restriction on the model.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size