Merge pull request #84 from HugoNjb/FindBestModel

anova.merMod error fix

R/find_best_model.lmerModLmerTest.R

@@ -1,71 +1,82 @@
-#' Returns the best combination of predictors for lmerTest objects.
-#'
-#' Returns the best combination of predictors for lmerTest objects.
-#'
-#' @param fit A merModLmerTest object.
-#' @param interaction Include interaction term.
-#' @param fixed Additional formula part to add at the beginning of
-#' each formula
-#' @param ... Arguments passed to or from other methods.
-#'
-#' @return list containing all combinations.
-#'
-#' @examples
-#' \dontrun{
-#' library(psycho)
-#' library(lmerTest)
-#'
-#' data <- standardize(iris)
-#' fit <- lmerTest::lmer(Sepal.Length ~ Sepal.Width + Petal.Length + (1|Species), data=data)
-#'
-#' best <- find_best_model(fit)
-#' best_formula <- best$formula
-#' best$table
-#'
-#' }
-#'
-#' @author \href{https://dominiquemakowski.github.io/}{Dominique Makowski}
-#'
-#' @importFrom stats update
-#' @import dplyr
-#'
-#' @method find_best_model lmerModLmerTest
-#' @export
-find_best_model.lmerModLmerTest <- function(fit, interaction=TRUE, fixed=NULL, ...) {
-
-  # Extract infos
-  combinations <- find_combinations(as.formula(get_formula(fit)), interaction = interaction, fixed = fixed)
-
-  # fit models
-  models <- c()
-  for (formula in combinations) {
-    newfit <- update(fit, formula)
-    models <- c(models, newfit)
-  }
-
-
-  # Model comparison
-  comparison <- as.data.frame(do.call("anova", models))
-  comparison$formula <- combinations
-
-
-  # Best model by criterion
-  best_aic <- dplyr::arrange_(comparison, "AIC") %>%
-    dplyr::select_("formula") %>%
-    head(1)
-  best_aic <- as.character(best_aic[[1]])
-
-  best_bic <- dplyr::arrange_(comparison, "BIC") %>%
-    dplyr::select_("formula") %>%
-    head(1)
-  best_bic <- as.character(best_bic[[1]])
-
-  by_criterion <- data.frame(formula = c(best_aic, best_bic), criterion = c("AIC", "BIC"))
-
-  # Best formula
-  best <- table(by_criterion$formula)
-  best <- names(best[which.max(best)])
-
-  best <- list(formula = best, by_criterion = by_criterion, table = comparison)
-  return(best)
-}
+#' Returns the best combination of predictors for lmerTest objects.
+#'
+#' Returns the best combination of predictors for lmerTest objects.
+#'
+#' @param fit A merModLmerTest object.
+#' @param interaction Include interaction term.
+#' @param fixed Additional formula part to add at the beginning of
+#' each formula
+#' @param ... Arguments passed to or from other methods.
+#'
+#' @return list containing all combinations.
+#'
+#' @examples
+#' \dontrun{
+#' library(psycho)
+#' library(lmerTest)
+#'
+#' data <- standardize(iris)
+#' fit <- lmerTest::lmer(Sepal.Length ~ Sepal.Width + Petal.Length + (1|Species), data=data)
+#'
+#' best <- find_best_model(fit)
+#' best_formula <- best$formula
+#' best$table
+#'
+#' }
+#'
+#' @author \href{https://dominiquemakowski.github.io/}{Dominique Makowski}
+#'
+#' @importFrom stats update
+#' @import dplyr
+#'
+#' @method find_best_model lmerModLmerTest
+#' @export
+find_best_model.lmerModLmerTest <- function(fit, interaction=TRUE, fixed=NULL, ...) {
+
+  # Extract infos
+  combinations <- find_combinations(as.formula(get_formula(fit)), interaction = interaction, fixed = fixed)
+
+
+  # Recreating the dataset without NA
+  dataComplete <- get_all_vars(fit)[complete.cases(get_all_vars(fit)), ]
+
+
+  # fit models
+  models <- c()
+  for (formula in combinations) {
+    newfit <- update(fit, formula, data = dataComplete)
+    models <- c(models, newfit)
+  }
+
+
+  # No warning messages for this part
+  options(warn = -1)
+
+  # Model comparison
+  comparison <- as.data.frame(do.call("anova", models))
+  comparison$formula <- combinations
+
+  # Re-displaying warning messages
+  options(warn = 0)
+
+
+  # Best model by criterion
+  best_aic <- dplyr::arrange_(comparison, "AIC") %>%
+    dplyr::select_("formula") %>%
+    head(1)
+  best_aic <- as.character(best_aic[[1]])
+
+  best_bic <- dplyr::arrange_(comparison, "BIC") %>%
+    dplyr::select_("formula") %>%
+    head(1)
+  best_bic <- as.character(best_bic[[1]])
+
+  by_criterion <- data.frame(formula = c(best_aic, best_bic), criterion = c("AIC", "BIC"))
+
+  # Best formula
+  best <- table(by_criterion$formula)
+  best <- names(best[which.max(best)])
+
+  best <- list(formula = best, by_criterion = by_criterion, table = comparison)
+  return(best)
+}