R/cov_matrix.R

In spmodel: Spatial Statistical Modeling and Prediction

#' Create a covariance matrix
#'
#' @param spcov_params A \code{spcov_params} object
#' @param dist_matrix A distance matrix (that has already been transformed for anisotropy)
#' @param randcov_params A \code{randcov_params} object
#' @param randcov_Zs Random effect design matrices
#' @param partition_matrix Partition matrix
#' @param M An M matrix for autoregressive models
#'
#' @return A covariance matrix
#'
#' @noRd
cov_matrix <- function(spcov_params, dist_matrix, randcov_params = NULL, randcov_Zs = NULL, partition_matrix = NULL, M = NULL, diagtol = 0) {

  # spatial
  if (is.null(M)) {
    cov_matrix_val <- spcov_matrix(spcov_params, dist_matrix, diagtol = diagtol)
  } else {
    cov_matrix_val <- spcov_matrix(spcov_params, dist_matrix, M)
  }


  # random effects
  if (!is.null(randcov_params)) {
    randcov_matrix_val <- randcov_matrix(randcov_params, randcov_Zs)
    cov_matrix_val <- cov_matrix_val + randcov_matrix_val
  }

  # partitioning
  if (!is.null(partition_matrix)) {
    cov_matrix_val <- cov_matrix_val * partition_matrix
  }

  # diag_add <- min(1e-4, 1e-4 * sum(spcov_params[["de"]], randcov_params))
  # diag(cov_matrix_val) <- diag(cov_matrix_val) + diag_add #
  # possibly needed for random effects stability with no ie
  cov_matrix_val
}


cov_matrix_cross <- function(spcov_params, dist_matrix_cross, randcov_params = NULL, randcov_Zs_cross = NULL, partition_matrix_cross = NULL) {

  # spatial
  spcov_params_ie <- spcov_params[["ie"]]
  spcov_params[["ie"]] <- 0
  cov_matrix_cross_val <- spcov_matrix(spcov_params, dist_matrix_cross)

  # random effects
  if (!is.null(randcov_params)) {
    randcov_matrix_cross_val <- randcov_matrix(randcov_params, randcov_Zs_cross)
    cov_matrix_cross_val <- cov_matrix_cross_val + randcov_matrix_cross_val
  }

  # partitioning
  if (!is.null(partition_matrix_cross)) {
    cov_matrix_cross_val <- cov_matrix_cross_val * partition_matrix_cross
  }

  cov_matrix_cross_val
}