mar_hmm_pn2pw: Transform multivariate autoregressive natural parameters to...

In longjess/hornsharkHMM: Functions For Analysing Horn Shark Data Using Hidden Markov Models

Transform multivariate autoregressive natural parameters to working parameters

Description

mu and phi do not need to be transformed, as there are no constraints. We only need to transform diagonal elements of sigma, since there are no constraints on the covariances. Include only the lower triangular and diagional elements of the sigma matrix, since covariance matrices must be symmetric.

Usage

mar_hmm_pn2pw(m, mu, sigma, gamma, phi, delta = NULL, stationary = TRUE)

Arguments

m	Number of states
mu	List of vectors of length m, means for white noise in each state dependent distribution
sigma	List of matrices of size m x m, covariance matrices for each state dependent distribution
gamma	Transition probabiilty matrix, size m x m
phi	List of k x (k x q) matrices, containing the autoregressive parameters. Each matrix corresponds to a state. The first k x k entries are the parameters for index i - 1, and so on up to index i - q.
delta	Optional, vector of length m containing initial distribution
stationary	Boolean, whether the HMM is stationary or not

Value

Vector of working parameters

longjess/hornsharkHMM documentation built on June 15, 2022, 11:32 p.m.