fit_multichannel_sequence_hmm: Fit a multichannel categorical hidden Markov model

In gp3sequences: Transparent Analysis of Ordered Categorical Sequences

Fit a multichannel categorical hidden Markov model

Description

Fits a finite-state, time-homogeneous HMM to two or more categorical channels under conditional independence of channels given the latent state. Latent states are statistical model states only.

Usage

fit_multichannel_sequence_hmm(
  data,
  n_states,
  channel_cols,
  sequence_id_col = "sequence_id",
  order_col = "sequence_order",
  symbol_levels = NULL,
  state_names = NULL,
  initial_probs = NULL,
  transition_probs = NULL,
  emission_probs = NULL,
  max_iter = 200L,
  tolerance = 1e-06,
  pseudocount = 1e-06,
  seed = 1L,
  keep_posteriors = FALSE
)

Arguments

data	Long-format multichannel sequence data.
n_states	Number of latent states.
channel_cols	Names of categorical observation channels.
sequence_id_col, order_col	Core sequence columns.
symbol_levels	Optional named list of symbol orders by channel.
state_names	Optional latent-state names.
initial_probs, transition_probs, emission_probs	Optional starting values.
max_iter	Maximum EM iterations.
tolerance	Relative log-likelihood tolerance.
pseudocount	Non-negative smoothing count.
seed	Reproducibility seed.
keep_posteriors	Retain final forward-backward results.

Value

An object of class gp3_multichannel_sequence_hmm.

Examples

multichannel <- data.frame(
  sequence_id = rep(paste0("s", 1:4), each = 5L),
  sequence_order = rep(1:5, times = 4L),
  action = c("A", "B", "C", "C", "D", "A", "B", "B", "C", "D",
             "D", "C", "B", "A", "A", "D", "C", "C", "B", "A"),
  context = rep(c("x", "x", "y", "y", "z"), times = 4L)
)
fit_multichannel_sequence_hmm(multichannel, 2L,
                              channel_cols = c("action", "context"),
                              max_iter = 5L, seed = 1L)

gp3sequences documentation built on Aug. 23, 2026, 5:10 p.m.