update_epsilon: Update the subject specific scaling parameters (epsilon)

In NewcastleCL/samplers: Particle Metropolis Within Gibbs

Update the subject specific scaling parameters (epsilon)

Description

Update the subject specific scaling parameter according to procedures outlined in P. H. Garthwaite, Y. Fan & S. A. Sisson (2016) Adaptive optimal scaling of Metropolis–Hastings algorithms using the Robbins–Monro process, Communications in Statistics - Theory and Methods, 45:17, 5098-5111, DOI: 10.1080/03610926.2014.936562

Usage

update_epsilon(epsilon, acc, p, i, d, alpha)

Arguments

epsilon	The scaling parameter for all subjects
acc	A boolean vector, TRUE if current sample != last sample
p	The target sample acceptance rate (0-1)
i	The current iteration for sampling
d	The number of parameters for the model
alpha	A hyperparameter for the epsilon tuning

Value

A vector with the new subject specific epsilon values

NewcastleCL/samplers documentation built on Feb. 12, 2024, 7:40 a.m.