Message/Author 

AHYOUNG posted on Wednesday, October 14, 2015  1:55 pm



I'm running a second order model but getting this warning message: THE LATENT VARIABLE COVARIANCE MATRIX (PSI) IS NOT POSITIVE DEFINITE...PROBLEM INVOLVING VARIABLE F1. The model is this: c@1; f1 BY u1u12; f2 BY u13u28; f3 BY u29u42; f4 BY u43u50; f5 BY u51u53; c BY f1* f2f5; I see that the residual for F1 is negative and the estimated correlation between c and f1 is over 1. The problem was solved when I fixed the residual for f1 as 0 by adding "f1@0;"in the code. But I would like to know if there is any other way to solve the problem besides this. Thanks for your help in advance! 


One reason this happens could be that the simple CFA structure (zero crossloadings) for the firstorder factors may cause too high firstorder factor correlations. 

Alice posted on Wednesday, October 14, 2015  3:33 pm



Dear Bengt O. Muthen, Will it be possible to suggest an example code that can be added to the above model that might help solve the problem? 


Use Modindices to check if there are some crossloadings or residual covariances among indicators that need to be added. 

Alice posted on Wednesday, October 14, 2015  6:47 pm



Thank you very much! 

Back to top 