Hi! For now, I've been doing measurement models on my 6 constructs. Most of the Fit Indices run well, but i simple can't "get rid off" the warning message: "The latent variable covariance matrix (PSI) is not positive definite...." - I do not have negative residuals - I think, i've already took my factor correlation >1 off (e.g., AUTSUP@1STRUTSUP@1; AUTSUP WITH STRUTSUP@1;) following your suggestion post: "Unless you have set the metric of the factor by freeing all factor loadings and fixing the factor variance to one, you are fixing the covariance to one not the correlation". The third correction alternative, i.e.., a linear dependency among more than 2 latent variables.." I think i shouldn't leave LV, as i just have 3 variables to regress on 17 items.. Thank you!
Hello! Trying to leave the warning message from "PSI" factor variance/covariance and after previous proceedings made to the model: 2 of the indicators >1 were fixed to 1. Output didn't showed negative variances. And an EFA didn't found "so high" correlations between factors: GEOMIN FACTOR CORRELATIONS (* significant at 5% level) 1 2 3 ________ ________ ________ 1 1.000 2 0.885* 1.000 3 0.746* 0.689* 1.000.
Afterwards, i made a 2nd Order Factor - but the warning "psi" message continues to pop up... Thank you! ps: Theory do not allowed me to run a 2 factor variable (during 1st Order Factor)