Help with factor analysis (latent variable model)

[Bob_Carpenter]

@andre.pfeuffer

is already in the Stan manual pg. 350.

I didn’t realize that was in there. I’m guessing Ben added it at some point.

[Johannes_Bjork]

@ldeschamps

// Correlation matrix of factors Ld = diag_pre_multiply(FS_sd, Rho);

Just of curiosity, why (in the context of your data) do you compute the above?

Thanks

[ldeschamps]

Hi!
This operation aims just to create a covariance matrix with estimated correlations between factors but diagonal fixed to one. This the Φ of classical factor analysis formulation. But I just realized that using direcly a constrained correlation matrix should work, without an unnecessary operation!