Fix problem with percentage_central argument in check_outliers() with MCD method

DESCRIPTION

@@ -1,7 +1,7 @@
 Type: Package
 Package: performance
 Title: Assessment of Regression Models Performance
-Version: 0.10.8.10
+Version: 0.10.8.11
 Authors@R: 
     c(person(given = "Daniel",
              family = "Lüdecke",
@@ -144,7 +144,7 @@ Suggests:
     withr (>= 3.0.0)
 Encoding: UTF-8
 Language: en-US
-RoxygenNote: 7.2.3.9000
+RoxygenNote: 7.3.1
 Roxygen: list(markdown = TRUE)
 Config/testthat/edition: 3
 Config/testthat/parallel: true

NEWS.md

@@ -23,6 +23,9 @@
 * `performance_score()` should no longer fail for models where scoring rules
   can't be calculated. Instead, an informative message is returned.
 
+* `check_outliers()` now properly accept the `percentage_central` argument when
+  using the `"mcd"` method.
+
 # performance 0.10.8
 
 ## Changes

R/check_heterogeneity_bias.R

@@ -31,9 +31,9 @@ check_heterogeneity_bias <- function(x, select = NULL, group = NULL) {
   if (insight::is_model(x)) {
     group <- insight::find_random(x, split_nested = TRUE, flatten = TRUE)
     if (is.null(group)) {
-      insight::format_error("Model is no mixed model. Please provide a mixed model, or a data frame and arguments `select` and `group`.")
+      insight::format_error("Model is no mixed model. Please provide a mixed model, or a data frame and arguments `select` and `group`.") # nolint
     }
-    data <- insight::get_data(x, source = "mf", verbose = FALSE)
+    my_data <- insight::get_data(x, source = "mf", verbose = FALSE)
     select <- insight::find_predictors(x, effects = "fixed", component = "conditional", flatten = TRUE)
   } else {
     if (inherits(select, "formula")) {
@@ -42,15 +42,15 @@ check_heterogeneity_bias <- function(x, select = NULL, group = NULL) {
     if (inherits(group, "formula")) {
       group <- all.vars(group)
     }
-    data <- x
+    my_data <- x
   }
 
-  unique_groups <- .n_unique(data[[group]])
+  unique_groups <- .n_unique(my_data[[group]])
   combinations <- expand.grid(select, group)
 
   result <- Map(function(predictor, id) {
     # demean predictor
-    d <- datawizard::demean(data, select = predictor, group = id, verbose = FALSE)
+    d <- datawizard::demean(my_data, select = predictor, group = id, verbose = FALSE)
 
     # get new names
     within_name <- paste0(predictor, "_within")

R/check_model.R

@@ -49,7 +49,7 @@
 #'
 #' @details For Bayesian models from packages **rstanarm** or **brms**,
 #' models will be "converted" to their frequentist counterpart, using
 #' [`bayestestR::bayesian_as_frequentist`](https://easystats.github.io/bayestestR/reference/convert_bayesian_as_frequentist.html).
 #' A more advanced model-check for Bayesian models will be implemented at a
 #' later stage.
 #'
@@ -77,7 +77,7 @@
 #' plots are helpful to check model assumptions, they do not necessarily indicate
 #' so-called "lack of fit", e.g. missed non-linear relationships or interactions.
 #' Thus, it is always recommended to also look at
 #' [effect plots, including partial residuals](https://strengejacke.github.io/ggeffects/articles/introduction_partial_residuals.html).
 #'
 #' @section Homogeneity of Variance:
 #' This plot checks the assumption of equal variance (homoscedasticity). The
@@ -184,14 +184,12 @@
   minfo <- insight::model_info(x, verbose = FALSE)
 
   ca <- tryCatch(
-    {
-      if (minfo$is_bayesian) {
-        suppressWarnings(.check_assumptions_stan(x, ...))
-      } else if (minfo$is_linear) {
-        suppressWarnings(.check_assumptions_linear(x, minfo, verbose, ...))
-      } else {
-        suppressWarnings(.check_assumptions_glm(x, minfo, verbose, ...))
-      }
+    if (minfo$is_bayesian) {
+      suppressWarnings(.check_assumptions_stan(x, ...))
+    } else if (minfo$is_linear) {
+      suppressWarnings(.check_assumptions_linear(x, minfo, verbose, ...))
+    } else {
+      suppressWarnings(.check_assumptions_glm(x, minfo, verbose, ...))
     },
     error = function(e) {
       NULL

R/check_model_diagnostics.R

@@ -94,29 +94,25 @@
   insight::check_if_installed("lme4")
 
   tryCatch(
-    {
-      if (inherits(model, "glmmTMB")) {
-        var_attr <- "condVar"
-        re <- .collapse_cond(lme4::ranef(model, condVar = TRUE))
-      } else {
-        var_attr <- "postVar"
-        re <- lme4::ranef(model, condVar = TRUE)
-      }
+    if (inherits(model, "glmmTMB")) {
+      var_attr <- "condVar"
+      re <- .collapse_cond(lme4::ranef(model, condVar = TRUE))
+    } else {
+      var_attr <- "postVar"
+      re <- lme4::ranef(model, condVar = TRUE)
     },
     error = function(e) {
-      return(NULL)
+      NULL
     }
   )
 
 
   se <- tryCatch(
-    {
-      suppressWarnings(lapply(re, function(.x) {
-        pv <- attr(.x, var_attr, exact = TRUE)
-        cols <- seq_len(dim(pv)[1])
-        unlist(lapply(cols, function(.y) sqrt(pv[.y, .y, ])))
-      }))
-    },
+    suppressWarnings(lapply(re, function(.x) {
+      pv <- attr(.x, var_attr, exact = TRUE)
+      cols <- seq_len(dim(pv)[1])
+      unlist(lapply(cols, function(.y) sqrt(pv[.y, .y, ])))
+    })),
     error = function(e) {
       NULL
     }
@@ -186,15 +182,15 @@
   n_params <- tryCatch(model$rank, error = function(e) insight::n_parameters(model))
 
   infl <- stats::influence(model, do.coef = FALSE)
-  resid <- as.numeric(insight::get_residuals(model))
+  model_resid <- as.numeric(insight::get_residuals(model))
 
-  std_resid <- tryCatch(stats::rstandard(model, infl), error = function(e) resid)
+  std_resid <- tryCatch(stats::rstandard(model, infl), error = function(e) model_resid)
 
   plot_data <- data.frame(
     Hat = infl$hat,
     Cooks_Distance = stats::cooks.distance(model, infl),
     Fitted = insight::get_predicted(model, ci = NULL),
-    Residuals = resid,
+    Residuals = model_resid,
     Std_Residuals = std_resid,
     stringsAsFactors = FALSE
   )
@@ -213,12 +209,10 @@
 
 .diag_ncv <- function(model, verbose = TRUE) {
   ncv <- tryCatch(
-    {
-      data.frame(
-        x = as.numeric(stats::fitted(model)),
-        y = as.numeric(stats::residuals(model))
-      )
-    },
+    data.frame(
+      x = as.numeric(stats::fitted(model)),
+      y = as.numeric(stats::residuals(model))
+    ),
     error = function(e) {
       NULL
     }
@@ -244,24 +238,22 @@
 .diag_homogeneity <- function(model, verbose = TRUE) {
   faminfo <- insight::model_info(model)
   r <- tryCatch(
-    {
-      if (inherits(model, "merMod")) {
-        stats::residuals(model, scaled = TRUE)
-      } else if (inherits(model, "gam")) {
-        stats::residuals(model, type = "scaled.pearson")
-      } else if (inherits(model, c("glmmTMB", "MixMod"))) {
-        sigma <- if (faminfo$is_mixed) {
-          sqrt(insight::get_variance_residual(model))
-        } else {
-          .sigma_glmmTMB_nonmixed(model, faminfo)
-        }
-        stats::residuals(model) / sigma
-      } else if (inherits(model, "glm")) {
-        ## TODO: check if we can / should use deviance residuals (as for QQ plots) here as well?
-        stats::rstandard(model, type = "pearson")
+    if (inherits(model, "merMod")) {
+      stats::residuals(model, scaled = TRUE)
+    } else if (inherits(model, "gam")) {
+      stats::residuals(model, type = "scaled.pearson")
+    } else if (inherits(model, c("glmmTMB", "MixMod"))) {
+      residual_sigma <- if (faminfo$is_mixed) {
+        sqrt(insight::get_variance_residual(model))
       } else {
-        stats::rstandard(model)
+        .sigma_glmmTMB_nonmixed(model, faminfo)
       }
+      stats::residuals(model) / residual_sigma
+    } else if (inherits(model, "glm")) {
+      ## TODO: check if we can / should use deviance residuals (as for QQ plots) here as well?
+      stats::rstandard(model, type = "pearson")
+    } else {
+      stats::rstandard(model)
     },
     error = function(e) {
       NULL