neg_log_lik_RE_logscale: negative log-likelihood for variance of random effect (call...

View source: R/random_effect_learner.R View source: R/random_effect_fit_functions.R

neg_log_lik_RE_logscaleR Documentation

negative log-likelihood for variance of random effect (call it V), on log-scale

Description

a ~ N(0, VR) for fixed correlation matrix R = USU' (of rank k <= n)

z ~ N(0, VI_k)

a = US^(1/2)z

y = a + e

e ~ N(0, diag(sigma2)), L = diag(1/sigma2)

Usage

neg_log_lik_RE_logscale(lV, SUtLY, SUtLUS)

Arguments

lV

is ln(V)

SUtLY

is S^(1/2)U'LY

SUtLUS

is S^(1/2)U'LUS^(1/2) (it must have an attribute called 'evd' that has its eigenvalue decomposition)


stephenslab/VEB.Boost documentation built on July 2, 2023, 1 p.m.