A negative variance/residual variance... PreviousNext
Mplus Discussion > Growth Modeling of Longitudinal Data >
Message/Author
 Ping Li posted on Wednesday, December 26, 2012 - 10:02 pm
Hi Linda,

I run a second-order lgc with two time points. My model is below:

MODEL:
LN1 BY LN11* LN12 LN13 ;
VT1 BY VT11* VT12 VT13 ;
LN1@1 VT1@1;

f1 BY LN1@1
VT1 (1);

f1@1;

LN2 BY LN21* LN22 LN23 ;
VT2 BY VT21* VT22 VT23 ;
LN2@1 VT2@1;

f2 BY LN2@1 VT2 (1);

f2@1;

[LN11 LN21] (2);
[ VT11 VT21] (3);
[LN12 LN22] (4);
[VT12 VT22] (5);
[LN13 LN23](6);
[VT13 VT23] (7);
[LN1 LN2](8);
[VT1 VT2](9)
i s |f1@0 f2@1;

Then I got an warning:

WARNING: THE LATENT VARIABLE COVARIANCE MATRIX (PSI) IS NOT POSITIVE
DEFINITE. THIS COULD INDICATE A NEGATIVE VARIANCE/RESIDUAL VARIANCE FOR A
LATENT VARIABLE,
CHECK THE TECH4 OUTPUT FOR MORE INFORMATION.
PROBLEM INVOLVING VARIABLE I.

Is it okay with negative variance? Thanks very much!

Ping
 Linda K. Muthen posted on Wednesday, December 26, 2012 - 10:09 pm
No, a negative variance means the model is not admissible. If it is very small and not significant, you can fix it at zero.

A linear model cannot be identified for only two time points. You need at least three.
 Ping Li posted on Thursday, December 27, 2012 - 6:13 pm
Thanks very much, Linda! Happy Holiday!
Back to top
Add Your Message Here
Post:
Username: Posting Information:
This is a private posting area. Only registered users and moderators may post messages here.
Password:
Options: Enable HTML code in message
Automatically activate URLs in message
Action: