update_Q: update the Q matrix element-wise

In zhenkewu/rewind: Reconstructing Etiology with Binary Decomposition

Description

This function updates Q matrix by Gibbs sampler (with option to do constrained updates within the identifiability constraint). Only used when Q is not known. NB: - do we need M here? do we modify M after collapsing partner machines. - need to speed up.

Usage

update_Q(Y, Q_old, H, z, t, mylist, p, theta, psi, constrained = FALSE)

Arguments

Y	binary data matrix
Q_old	the Q matrix from the last scan in Gibbs sampler (of dimension M by L)
H	matrix of latent state profiles, rows for N subjects, columns for M latent states
z	a vector of individual cluster indicators
t	the number of clusters at a particular iteration
mylist	the ordered list that keeps the cluster ids in increasing order
p	Latent state prevalences; a vector of length identical to the columns of H.
theta, psi	True and false positive rates. Both are vectors of length L
constrained	Default to FALSE; Set to TRUE if doing constrained Gibbs sampling within identifiability constraints.

Value

A Q matrix with all its elements updated.

zhenkewu/rewind documentation built on Sept. 9, 2020, 3:40 p.m.