Correlation i with s PreviousNext
Mplus Discussion > Growth Modeling of Longitudinal Data >
 Stefan Koenig posted on Monday, February 19, 2018 - 7:34 am

in an LGCM on physical endurance measured with motor tests Mplus reports correlation coefficients (i with s) which are far beyond -1 and +1. I do not understand these estimations.

Thank you and best regards
 Bengt O. Muthen posted on Monday, February 19, 2018 - 4:40 pm
That says that you have a poor model that should be re-specified in some way. Perhaps a random s is not needed for instance.
 Stefan Koenig posted on Monday, February 19, 2018 - 10:54 pm
Indeed, some model fits are poor. However I thought this correlation has to be between -1 and +1. In this case it is -17,...

Thank you
 Stefan Koenig posted on Tuesday, February 20, 2018 - 2:25 am
Postscript: Is it an issue of standardization in the Output command?
 Bengt O. Muthen posted on Tuesday, February 20, 2018 - 2:27 pm
The variances and the covariances of i and s are freely estimated. When standardized the estimates may give rise to correlations greater than 1 in absolute value, that is, the covariance matrix for i and s is not positive definite. There are ways to enforce pos def but I think it is valuable to see such non-pos def to realize that the model is not good and modify it.
Back to top
Add Your Message Here
Username: Posting Information:
This is a private posting area. Only registered users and moderators may post messages here.
Options: Enable HTML code in message
Automatically activate URLs in message