multi_HMM_DM: Fitting a Hidden Markov Model via the Direct Maximisation
In pneff93/HMM: Fitting a Hidden Markov Model

Description Usage Arguments Details Value

Estimation of the transition probabilites, the initial state probabilites and the hidden state parameters of a Hidden Markov Model by using the Direct Maximisation of the global log-likelihood.

1	multi_HMM_DM(x, theta, m, L1, L2, L3, L4, L5)

`x`	a sample of a Hidden Markov Model
`theta`	initial parameters for the estimation of the likelihood parameters. See details for more information.
`m`	the number of states
`L1`	likelihood of the first hidden state
`L2`	likelihood of the second hidden state
`L3`	optional. likelihood of the third hidden state
`L4`	optional. likelihood of the 4th hidden state
`L5`	optional. likelihood of the 5th hidden state

This function estimates the Hidden Markov states by maximising the normalized log-likelihood of the forward propabilities. Due to the fact that both the Gamma matrix as well as the Delta vector have some constraints, the function first applies some restrictions and then uses the base-R maximisation to gain the most likely variables.

This function is able to calculate with multiple Theta values for the individual likelihoods. For each likelihood the right number initial starting parameter has to be set in order to compute the estimation of the corresponding Thetas. For each Likelihood the starting values must be in the format of a vector, which is then saved as a list element.

e.g.: theta[[i]] <- c(parameter1, parameter2, ...)

The function then extracts the right number of parameters per likelihood and optimizes the values.

The estimated parameters are rounded by 3 decimals and returned in a list.

pneff93/HMM documentation built on Oct. 26, 2019, 8:16 a.m.