Convergence problems with factor loadings in multivariante autoregressive model

Modeling
, , ,
5

For what I understood \Sigma=LL^T+\Psi where L is the matrix of factor loadings with the constrains from @rfarouni blog. In Help with factor analysis (latent variable model) they used multi_normal_cholesky for the factor scores, not for the factor loadings. The Ld in their code is the product between the factor scores sd and the factor scores correlation, whereas the L in my code is the factor loadings, what is defined as lambda in their code. So you are suggesting to write

model{
    multi_normal_cholesky(r + diag_matrix(a) * Y[t-1], L)
}

In this case I don’t know what to do with psi

Or

transformed parameters{
    L_s = diag_pre_multiply(psi, L)
}
model {
    multi_normal_cholesky(r + diag_matrix(a) * Y[t-1], L_s)
}

But L is not from a correlation matrix

I am sure I am missing something conceptually.

And thanks a lot for your help,
Alfonso