I'm running a multigroup comparison of a model that has observed variables and a latent interaction at Time 1, predicting observed variables at Time 2.
In the unconstrained initial model I obtain covariance coefficients (WITH statements) larger than 1 (in both unstandardized and standardized coefficients). Is this normal?
I'm not getting any error/warning message. I checked the variances and they are not negative, and the TECH4 and TECH1 outputs are not giving me that much info about potential mistakes.
It could be . Note the definition of of Cov(x,y)=E((x-mean(x))(y-mean(y)) =
Exy-mean(x)*mean(y) There is no Squared term as there is in variance i.e.
Var(x)=E(x-Ex)^2 that squared is what makes everything positive for variance.
Best, D. Booth we would have to see the data to be sure
David Eugene Booth , she said she gets also a standardized cov (=correlation) > 1. I would guess that this is not possible and indicates a identification problem (heywood case). Correct?
Best,
Holger
Holger Steinmetz You are,of course, correct. We perhaps need to see code and data to help. Best, David Booth
- Badges
- Science topic
- Similar topics
- Chemistry