Do we need to avoid a saturated model?
Posted on

Forums
Dear Dr. Cheung,
I used Lavaan to specify my model. Do I need to declare the correlations among the mediator variables besides the correlations among the independent variables?
After adding all paths to the model, I got a perfect fit model, which will be of little use statistically.
Thanks,
Xin
Dear Xin,
Dear Xin,
Many researchers prefer to correlate the mediators. If it is a saturated model, the focus is on the parameter estimates and their standard errors (or confidence intervals).
Best,
Mike
Log in or register to post comments
In reply to Dear Xin, by Mike Cheung
Dear Dr. Cheung,
Dear Dr. Cheung,
Thanks for your answer in detail. I appreciate it!
Best,
Xin
Log in or register to post comments