R/mse.R

In Redha-ALLICHE/clusterwiselmm: Apply geometrical clusterwise to linear multi-level models (lmm)

#' MSE error
#' @description Compute cost function for clusterwise lmm that is the mean squared error between
#' the prediction and the target variable
#'
#' @param prediction vector containing the response of the clusterwise lmm to the data.
#' @param target vector containing the target variable.
#' @param residuals logical; if TRUE, returns the squared of residuals instead of the MSE
#'
#' @return mean squared error between prediction and target
#'

mse <- function(prediction, target, residuals = FALSE) {
  if(length(prediction)==1) {
    if(is.infinite(prediction))
      return(Inf)
  }
  if (length(prediction) != length(target)) {
    stop("prediction and target must have the same length")
  }

  if (residuals) {
    errors = (prediction - target) ^ 2
  }

  else{
    errors = mean((prediction - target) ^ 2)
  }
  return(errors)
}