R/ols-error-messages.R

In olsrr: Tools for Building OLS Regression Models

check_model <- function(model) {
  if (!any(class(model) == "lm")) {
    model_name <- deparse(substitute(model))
    stop(paste0("`", model_name, "` must be an object of class `lm`."), call. = FALSE)
  }
}

check_order <- function(n, max_order) {
  if (max_order > n) {
    stop("Maximum subset order should be less than or equal to the number of predictors in the specified model.", call. = FALSE)
  }
}

check_logic <- function(logic) {

  lval        <- is.logical(logic)
  logic_name  <- deparse(substitute(logic))

  if (lval != TRUE) {
    stop(paste0("`", logic_name, '` must be a flag (TRUE or FALSE).'), call. = FALSE)
  }

}

check_options <- function(option) {

  default_options <- c("none", "bonferroni", "sidak", "holm")
  valid           <- any(default_options == option)
  option_class    <- class(option)
  option_name     <- deparse(substitute(option))

  if (valid != TRUE) {
    err_message <- paste0("`", option_name, "` can take the following values only:\n* `none`\n* `bonferroni`\n* `sidak`\n* `holm`")
    stop(err_message, call. = FALSE)
  }
}


check_values <- function(value, lower, upper) {

  valid       <- (value >= lower && value <= upper)
  value_class <- class(value)
  value_name  <- deparse(substitute(value))

  if (valid != TRUE) {
    err_message <- paste0("`", value_name, "` must be between ", lower, " and ", upper, ".")
    stop(err_message, call. = FALSE)
  }

}

check_npredictors <- function(model, min) {

  n <- length(coefficients(model))

  if (n < min) {
    err_message <- paste0("For stepwise selection, model should include at least `", min, "` predictors.")
    stop(err_message, call. = FALSE)
  }

}


check_lfit <- function(model) {

  n     <- length(coefficients(model))
  preds <- n - 1

  if (n > 2) {
    err_message <- paste0("Lack of fit F test is available only for models with a single predictor.")
    stop(err_message, call. = FALSE)
  }

}

check_modelvars <- function(model, vars) {

  fmla  <- deparse(formula(model))
  k     <- vars %in% names(model$coefficients)
  wvars <- vars[!k]
  nvars <- length(wvars)

  if (nvars > 0) {
    err_message <- paste0("Model `(", fmla, ")` should include the predictor(s) `", paste0(wvars, collapse = ", "), "`.")
    stop(err_message, call. = FALSE)
  }

}