- I conducted SEM using Lavaan on R with a measurement model, regressions, and categorical predictors. To improve model fit, I did item parceling.
- When I run this model, I run into an issue with no positive definite. When running a previous model (not the one from the previous point) using missing = FIML when I had continuous outcomes, it solved the issue, but with categorical outcomes I can’t do that. Google and ChatGPT have pointed to using missing = WLSMV which SHOULD not delete list wise and work with categorical outcomes, but the N is still much lower than what I know it should be compared to that other model I ran with FIML.
- Does anyone know of strategies to help solve this issue? Thanks in advance 🙂
Tried using missing = WLSMV but that still doesn’t fix the issue and I can tell that much data is missing.