Marginal within-biomarker variance under the factor model

The Model 2a factor construction is par[i,k,p] = mu[k,p] + w[i,k,p] + lambda[k,p] * f[i,p], so the marginal variance of parameter p for biomarker k is the within-biomarker variance of w plus the squared loading: $$\mathrm{Var}(par_{k,p}) = (\mathrm{prec.par}_k^{-1})_{pp} + \lambda_{k,p}^2.$$ This pure helper returns that marginal variance vector for one biomarker, for a single MCMC draw.

Usage

marginal_var_2a(prec_par_k, lambda_k)

Arguments

prec_par_k

A P x P precision matrix for biomarker k (the Wishart node prec.par[k,,]).

lambda_k

A length-P numeric vector of loadings for biomarker k.

Value

A length-P numeric vector of marginal variances.