DHARMa implementation for new check_residuals() function

DESCRIPTION

@@ -1,7 +1,7 @@
 Type: Package
 Package: performance
 Title: Assessment of Regression Models Performance
-Version: 0.10.9.6
+Version: 0.10.9.8
 Authors@R: 
     c(person(given = "Daniel",
              family = "Lüdecke",
@@ -70,9 +70,8 @@ Depends:
     R (>= 3.6)
 Imports:
     bayestestR (>= 0.13.2),
-    insight (>= 0.19.8),
+    insight (>= 0.19.9),
     datawizard (>= 0.9.1),
-    methods,
     stats,
     utils
 Suggests:
@@ -91,6 +90,7 @@ Suggests:
     correlation,
     cplm,
     dbscan,
+    DHARMa,
     estimatr,
     fixest,
     flextable, 

NAMESPACE

@@ -55,9 +55,11 @@ S3method(check_heteroscedasticity,default)
 S3method(check_homogeneity,afex_aov)
 S3method(check_homogeneity,default)
 S3method(check_homogeneity,htest)
+S3method(check_model,DHARMa)
 S3method(check_model,brmsfit)
 S3method(check_model,default)
 S3method(check_model,model_fit)
+S3method(check_model,performance_simres)
 S3method(check_model,stanreg)
 S3method(check_multimodal,data.frame)
 S3method(check_multimodal,numeric)
@@ -71,6 +73,7 @@ S3method(check_normality,lmerModLmerTest)
 S3method(check_normality,merMod)
 S3method(check_normality,numeric)
 S3method(check_outliers,BFBayesFactor)
+S3method(check_outliers,DHARMa)
 S3method(check_outliers,character)
 S3method(check_outliers,data.frame)
 S3method(check_outliers,default)
@@ -87,11 +90,13 @@ S3method(check_outliers,meta)
 S3method(check_outliers,metabin)
 S3method(check_outliers,metagen)
 S3method(check_outliers,numeric)
+S3method(check_outliers,performance_simres)
 S3method(check_outliers,rma)
 S3method(check_outliers,rma.uni)
 S3method(check_outliers,rq)
 S3method(check_outliers,rqs)
 S3method(check_outliers,rqss)
+S3method(check_overdispersion,DHARMa)
 S3method(check_overdispersion,default)
 S3method(check_overdispersion,fixest)
 S3method(check_overdispersion,fixest_multi)
@@ -103,11 +108,15 @@ S3method(check_overdispersion,model_fit)
 S3method(check_overdispersion,negbin)
 S3method(check_overdispersion,negbinirr)
 S3method(check_overdispersion,negbinmfx)
+S3method(check_overdispersion,performance_simres)
 S3method(check_overdispersion,poissonirr)
 S3method(check_overdispersion,poissonmfx)
 S3method(check_predictions,BFBayesFactor)
 S3method(check_predictions,default)
 S3method(check_predictions,lme)
+S3method(check_residuals,DHARMa)
+S3method(check_residuals,default)
+S3method(check_residuals,performance_simres)
 S3method(check_singularity,MixMod)
 S3method(check_singularity,clmm)
 S3method(check_singularity,cpglmm)
@@ -123,6 +132,9 @@ S3method(check_sphericity,default)
 S3method(check_sphericity,mlm)
 S3method(check_symmetry,htest)
 S3method(check_symmetry,numeric)
+S3method(check_zeroinflation,DHARMa)
+S3method(check_zeroinflation,default)
+S3method(check_zeroinflation,performance_simres)
 S3method(cronbachs_alpha,data.frame)
 S3method(cronbachs_alpha,matrix)
 S3method(cronbachs_alpha,parameters_pca)
@@ -261,10 +273,12 @@ S3method(plot,check_model)
 S3method(plot,check_normality)
 S3method(plot,check_outliers)
 S3method(plot,check_overdisp)
+S3method(plot,check_residuals)
 S3method(plot,check_sphericity)
 S3method(plot,compare_performance)
 S3method(plot,performance_pp_check)
 S3method(plot,performance_roc)
+S3method(plot,performance_simres)
 S3method(plot,test_likelihoodratio)
 S3method(plot,test_performance)
 S3method(print,binned_residuals)
@@ -283,7 +297,9 @@ S3method(print,check_normality_binom)
 S3method(print,check_outliers)
 S3method(print,check_outliers_metafor)
 S3method(print,check_outliers_metagen)
+S3method(print,check_outliers_simres)
 S3method(print,check_overdisp)
+S3method(print,check_residuals)
 S3method(print,check_sphericity)
 S3method(print,check_symmetry)
 S3method(print,check_zi)
@@ -302,6 +318,7 @@ S3method(print,performance_pcp)
 S3method(print,performance_pp_check)
 S3method(print,performance_roc)
 S3method(print,performance_score)
+S3method(print,performance_simres)
 S3method(print,r2_bayes)
 S3method(print,r2_generic)
 S3method(print,r2_loo)
@@ -535,6 +552,7 @@ export(check_outliers)
 export(check_overdispersion)
 export(check_posterior_predictions)
 export(check_predictions)
+export(check_residuals)
 export(check_singularity)
 export(check_sphericity)
 export(check_sphericity_bartlett)
@@ -588,6 +606,7 @@ export(r2_tjur)
 export(r2_xu)
 export(r2_zeroinflated)
 export(rmse)
+export(simulate_residuals)
 export(test_bf)
 export(test_likelihoodratio)
 export(test_lrt)

NEWS.md

@@ -4,10 +4,35 @@
 
 * Rudimentary support for models of class `serp` from package *serp*.
 
+## New functions
+
+* `simulate_residuals()` and `check_residuals()`, to simulate and check residuals
+  from generalized linear (mixed) models. Simulating residuals is based on the
+  DHARMa package, and objects returned by `simulate_residuals()` inherit from
+  the `DHARMa` class, and thus can be used with any functions from the *DHARMa*
+  package. However, there are also implementations in the *performance* package,
+  such as `check_overdispersion()`, `check_zeroinflation()`, `check_outliers()`
+  or `check_model()`.
+
+* Plots for `check_model()` have been improved. The Q-Q plots are now based
+  on simulated residuals from the DHARMa package for non-Gaussian models, thus
+  providing more accurate and informative plots. The half-normal QQ plot for
+  generalized linear models can still be obtained by setting the new argument
+  `residual_type = "normal"`.
+
+* Following functions now support simulated residuals (from `simulate_residuals()`)
+  resp. objects returned from `DHARMa::simulateResiduals()`:
+  - `check_overdispersion()`
+  - `check_zeroinflation()`
+  - `check_outliers()`
+  - `check_model()`
+
 ## General
 
 * Improved error messages for `check_model()` when QQ-plots cannot be created.
 
+* `check_distribution()` is more stable for possibly sparse data.
+
 ## Bug fixes
 
 * Fixed issue in `check_normality()` for t-tests.

R/check_distribution.R

@@ -77,7 +77,7 @@
   } else {
     x <- stats::residuals(model)
   }
-  dat <- .extract_features(x)
+  dat <- .extract_features(x, "residuals")
 
   dist_residuals <- as.data.frame(t(stats::predict(classify_distribution, dat, type = "prob")))
 
@@ -88,7 +88,7 @@
     dummy_factors = FALSE,
     preserve_levels = TRUE
   )
-  dat <- .extract_features(x)
+  dat <- .extract_features(x, "response")
 
   dist_response <- as.data.frame(t(stats::predict(classify_distribution, dat, type = "prob")))
 
@@ -189,15 +189,27 @@
 
 # utilities -----------------------------
 
-.extract_features <- function(x) {
+.extract_features <- function(x, type = NULL) {
   # validation check, remove missings
   x <- x[!is.na(x)]
 
+  # this might fail, so we wrap in ".safe()"
+  map_est <- .safe(mean(x) - as.numeric(bayestestR::map_estimate(x, bw = "nrd0")))
+
+  if (is.null(map_est)) {
+    map_est <- mean(x) - datawizard::distribution_mode(x)
+    msg <- "Could not accurately estimate the mode."
+    if (!is.null(type)) {
+      msg <- paste(msg, "Predicted distribution of the", type, "may be less accurate.")
+    }
+    insight::format_alert(msg)
+  }
+
   data.frame(
     SD = stats::sd(x),
     MAD = stats::mad(x, constant = 1),
     Mean_Median_Distance = mean(x) - stats::median(x),
-    Mean_Mode_Distance = mean(x) - as.numeric(bayestestR::map_estimate(x, bw = "nrd0")),
+    Mean_Mode_Distance = map_est,
     SD_MAD_Distance = stats::sd(x) - stats::mad(x, constant = 1),
     Var_Mean_Distance = stats::var(x) - mean(x),
     Range_SD = diff(range(x)) / stats::sd(x),