R/ProbMatrixQ.R

In GenMarkov: Multivariate Markov Chains

#' Probability Transition Matrices
#'
#' Calculate Probability Transition Matrices (Ching (2002))
#'
#' @param s numerical matrix with categorical data sequences
#'
#' @return numerical array with probability transition matrices
#' @keywords internal
#' @keywords internal
#' @noRd
#'
#' @examples
#' data(stockreturns)
#' s <- cbind(stockreturns$sp500, stockreturns$djia)
#' ProbMatrixQ(s)
ProbMatrixQ <- function(s) {
  n <- nrow(s)
  m1 <- max(s, na.rm = TRUE)
  m0 <- min(s, na.rm = TRUE)
  num_cols <- ncol(s)
  number_comb <- num_cols * num_cols

  s_prev <- s[-nrow(s), ]
  s_next <- s[-1, ]

  column_combinations <- expand.grid(1:num_cols, 1:num_cols)

  all_list <- lapply(1:nrow(column_combinations), function(i) {
    selected_columns <- column_combinations[i, 2:1]
    combined_data <- data.frame(s_next[, selected_columns[, 1]], s_prev[, selected_columns[, 2]])
    return(combined_data)
  })

  q <- lapply(
    all_list,
    FUN = function(z) {
      prop.table(table(z), margin = 2)
    }
  )

  return(array(unlist(q), dim = c(m1, m1, number_comb)))
}