permuteMCMC: Relabel MCMC output based on Permutation Matrix
In baruuum/relabelKL: Relabeling MCMC Output

permuteMCMC

R Documentation

Relabel MCMC output based on Permutation Matrix

Description

Relabel MCMC output from finite mixture models using the permutations found by some algorithm

Usage

permuteMCMC(x, perms, what = "cols")

Arguments

`x`	an array of MCMC samples, where the first index has to be equal to the number of posterior samples, `S`, and the last index equal to the number of latent classes/extreme types, `K`.
`perms`	a `S` times `K` matrix of permutations needed to transformed the raw MCMC output to the permuted output, e.g., the `perms` element from running `relabelMCMC`.
`what`	one of "rows", "cols," "both", or "matrix." Determines what dimensions are permuted. For two-dimensional arrays with unequal dimensions (i.e., rectangular matrices), the columns are always permuted.

Details

Given a permutation that resolves the label-switching, this function applies the mapping to another object. If the object to permute is a three-dimensional array of dimension S\*N\*K, the function will treat S as the number of posterior draws and will permute either the columns or the rows of the sub-array which is of dimensions N*K. If x is a matrix, the function assumes that nrow(x) == S and will permute the columns. When the what argument is set to "matrix," it is assumed that the entered array is of dimension S\*K\*L\*L, and permutes the whole L\*L matrix across the K classes for each posterior draw.