I have two variables (one data, one parameter) which should be upper or lower triangular matrices, but they aren’t Cholesky factors of anything. Is it ok to use the
cholesky_factor_cov data type for this?
edit if so, can a prior be set using
cholesky_factor_cov tri; to_vector(tri) ~ beta(a, b);
or would one need to loop through each element?
I was worried about declaring a
matrix parameter and the other triangle being unconstrained for estimation (and generally slowing things down).
I searched the forum and found only hints, but I wanted to confirm from an implementation perspective this works.