Check_model in version 0.11.0 no longer produces qq plot residuals

DESCRIPTION

@@ -1,7 +1,7 @@
 Type: Package
 Package: performance
 Title: Assessment of Regression Models Performance
-Version: 0.11.0.3
+Version: 0.11.0.4
 Authors@R: 
     c(person(given = "Daniel",
              family = "Lüdecke",

NEWS.md

@@ -15,6 +15,11 @@
   the usual style as for other models and no longer returns plots from
   `bayesplot::pp_check()`.
 
+## Bug fixes
+
+* `check_model()` now falls back on normal Q-Q plots when a model is not supported
+  by the DHARMa package and simulated residuals cannot be calculated.
+
 # performance 0.11.0
 
 ## New supported models

R/check_homogeneity.R

@@ -70,14 +70,14 @@
     )
 
     if (is.null(check)) {
       insight::print_color("'check_homogeneity()' cannot perform check for normality. Please specify the 'method'-argument for the test of equal variances.\n", "red")
       return(NULL)
     }
 
     method <- ifelse(check < 0.05, "fligner", "bartlett")
   }
 
   if (method == "fligner") {
     r <- stats::fligner.test(f, data = insight::get_data(x, verbose = FALSE))
     p.val <- r$p.value
   } else if (method == "bartlett") {
@@ -117,7 +117,7 @@
   } else if (x < 0.05) {
     insight::print_color(sprintf("Warning: Variances differ between groups (%s, p = %.3f).\n", method.string, x), "red")
   } else {
-    insight::print_color(sprintf("OK: There is not clear evidence for different variances across groups (%s, p = %.3f).\n", method.string, x), "green")
+    insight::print_color(sprintf("OK: There is not clear evidence for different variances across groups (%s, p = %.3f).\n", method.string, x), "green") # nolint
   }
   invisible(x)
 }
@@ -146,18 +146,18 @@
     insight::format_error("Levene test is only aplicable to ANOVAs with between-subjects factors.")
   }
 
-  data <- x$data$long # Use this to also get id column
+  long_data <- x$data$long # Use this to also get id column
   dv <- attr(x, "dv")
   id <- attr(x, "id")
   between <- names(attr(x, "between"))
   is_covar <- vapply(attr(x, "between"), is.null, logical(1))
 
-  ag_data <- stats::aggregate(data[, dv], data[, c(between, id)], mean)
+  ag_data <- stats::aggregate(long_data[, dv], long_data[, c(between, id)], mean)
   colnames(ag_data)[length(c(between, id)) + 1] <- dv
 
   if (any(is_covar)) {
     insight::format_alert(
       "Levene's test is not appropriate with quantitative explanatory variables. Testing assumption of homogeneity among factor groups only."
     )
     # ## TODO maybe add as option?
     # warning("Testing assumption of homogeneity on residualzied data among factor groups only.", call. = FALSE)

R/check_model.R

@@ -60,7 +60,7 @@
 #'
 #' @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.
 #'
@@ -88,7 +88,7 @@
 #' 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
@@ -213,6 +213,15 @@
   if (is.null(residual_type)) {
     residual_type <- ifelse(minfo$is_linear && !minfo$is_gam, "normal", "simulated")
   }
+
+  # catch models/families not supported by DHARMa - we need to add more
+  # exceptions here as they appear, but for now, `check_model()` also
+  # automatically falls back to normal Q-Q plot for all models not supported
+  # by DHARMa
+  if (minfo$family %in% c("quasipoisson", "quasibinomial")) {
+    residual_type <- "normal"
+  }
+
   # set default for detrend
   if (missing(detrend)) {
     detrend <- residual_type == "normal"
@@ -242,8 +251,8 @@
   }
 
   # did Q-Q plot work with simulated residuals?
-  if (verbose && is.null(assumptions_data$QQ) && residual_type == "simulated") {
-    insight::format_warning(paste0(
+  if (is.null(assumptions_data$QQ) && residual_type == "simulated") {
+    insight::format_alert(paste0(
       "Cannot simulate residuals for models of class `",
       class(x)[1],
       "`. Please try `check_model(..., residual_type = \"normal\")` instead."

R/check_predictions.R

@@ -536,7 +536,7 @@
   # validation check, for mixed models, where re.form = NULL (default) might fail
   if (is.null(out) && insight::is_mixed_model(object) && !isTRUE(is.na(re_formula))) {
     if (verbose) {
-      insight::format_warning(
+      insight::format_alert(
         paste0(
           "Failed to compute posterior predictive checks with `re_formula=",
           deparse(re_formula),

R/cronbachs_alpha.R

@@ -40,7 +40,7 @@
   # we need at least two columns for Cronach's Alpha
   if (is.null(ncol(.data)) || ncol(.data) < 2) {
     if (verbose) {
-      insight::format_warning("Too few columns in `x` to compute Cronbach's Alpha.")
+      insight::format_alert("Too few columns in `x` to compute Cronbach's Alpha.")
     }
     return(NULL)
   }
@@ -68,7 +68,7 @@
     pca_data <- attr(x, "data")
     if (is.null(pca_data)) {
       if (verbose) {
-        insight::format_warning("Could not find data frame that was used for the PCA.")
+        insight::format_alert("Could not find data frame that was used for the PCA.")
       }
       return(NULL)
     }

inst/WORDLIST

@@ -25,6 +25,7 @@ Bollen's
 Bortz
 Breunig
 Breusch
+BRM
 Bryk
 Bundock
 Burnham

tests/testthat/test-check_model.R

@@ -79,5 +79,33 @@ test_that("`check_model()` warnings for zero-infl", {
     art ~ fem + mar + kid5 + ment | kid5 + phd,
     data = bioChemists
   )
-  expect_message(expect_warning(check_model(model, verbose = TRUE), regex = "Cannot simulate"), regex = "Homogeneity")
+  expect_message(expect_message(check_model(model, verbose = TRUE), regex = "Cannot simulate"), regex = "Homogeneity")
+})
+
+
+test_that("`check_model()` no warnings for quasipoisson", {
+  skip_if_not_installed("datawizard")
+  set.seed(250419)
+  # Generate random x values
+  x <- rnorm(
+    n = 500,
+    mean = 5,
+    sd = 2
+  )
+  # Generate y values y = 5x + e
+  y <- 5 * x + rnorm(
+    n = 500,
+    mean = 5,
+    sd = 2
+  )
+  # Generate z as offset
+  z <- runif(500, min = 0, max = 6719)
+  mock_data <- data.frame(x, y, z) |>
+    # both should be whole numbers since they're counts
+    datawizard::data_modify(y = round(y), z = round(z)) |>
+    datawizard::data_filter(x >= 0, y >= 0)
+  # Run model
+  model1 <- glm(y ~ x + offset(log(z)), family = "quasipoisson", data = mock_data)
+  expect_message(check_model(model1, verbose = TRUE), regex = "Not enough")
+  expect_silent(check_model(model1))
 })

tests/testthat/test-cronbachs_alpha.R

@@ -3,16 +3,16 @@ test_that("cronbachs_alpha, data frame", {
   x <- mtcars[, c("cyl", "gear", "carb", "hp")]
   expect_equal(cronbachs_alpha(x), 0.09463206, tolerance = 1e-3)
 })
+
 test_that("cronbachs_alpha", {
-  expect_warning(expect_null(cronbachs_alpha(mtcars[1])))
+  expect_message(expect_null(cronbachs_alpha(mtcars[1])), regex = "Too few")
 })
 
-
 test_that("cronbachs_alpha, principal_components", {
   skip_if_not_installed("parameters", minimum_version = "0.21.3")
   pca <- parameters::principal_components(mtcars[, c("cyl", "gear", "carb", "hp")], n = 2)
   expect_equal(cronbachs_alpha(pca, verbose = FALSE), c(PC1 = 0.1101384), tolerance = 1e-3)
-  expect_warning(cronbachs_alpha(pca))
+  expect_message(cronbachs_alpha(pca), regex = "Too few")
 
   pca <- parameters::principal_components(mtcars[, c("cyl", "gear", "carb", "hp")], n = 1)
   expect_equal(cronbachs_alpha(pca, verbose = FALSE), c(PC1 = 0.09463206), tolerance = 1e-3)
@@ -25,7 +25,6 @@ test_that("cronbachs_alpha, principal_components", {
   expect_equal(cronbachs_alpha(pca), c(PC1 = 0.4396, PC2 = -1.44331), tolerance = 1e-3)
 })
 
-
 test_that("cronbachs_alpha, matrix", {
   m <- as.matrix(mtcars[c("cyl", "gear", "carb", "hp")])
   expect_equal(cronbachs_alpha(m), 0.09463206, tolerance = 1e-3)