Should we add allow.new.levels=TRUE by default? (#987)

* Should we add `allow.new.levels=TRUE` by default?
Fixes easystats/modelbased#285

* news

* wordlist

DESCRIPTION

@@ -1,7 +1,7 @@
 Type: Package
 Package: insight
 Title: Easy Access to Model Information for Various Model Objects
-Version: 1.0.0.5
+Version: 1.0.0.6
 Authors@R:
     c(person(given = "Daniel",
              family = "Lüdecke",

NEWS.md

@@ -19,6 +19,9 @@
 
 * Fixed issue with `as.numeric()` method for `get_sigma()`.
 
+* `get_datagrid()` now only returns valid levels when `include_random = TRUE`
+  and group-level factor in random effects is numeric.
+
 # insight 1.0.0
 
 ## Breaking changes

R/get_datagrid.R

@@ -518,14 +518,20 @@ get_datagrid.default <- function(x,
 
   # Drop random factors
   random_factors <- find_random(x, flatten = TRUE)
-  if (isFALSE(include_random) && !is.null(random_factors)) {
-    keep <- c(find_predictors(x, effects = "fixed", flatten = TRUE), response)
-    if (!is.null(keep)) {
-      if (all(by != "all")) {
-        keep <- c(keep, by[by %in% random_factors])
-        random_factors <- setdiff(random_factors, by)
+  if (!is.null(random_factors)) {
+    if (isFALSE(include_random)) {
+      # drop random factors, if these should not be included
+      keep <- c(find_predictors(x, effects = "fixed", flatten = TRUE), response)
+      if (!is.null(keep)) {
+        if (all(by != "all")) {
+          keep <- c(keep, by[by %in% random_factors])
+          random_factors <- setdiff(random_factors, by)
+        }
+        data <- data[colnames(data) %in% keep]
       }
-      data <- data[colnames(data) %in% keep]
+    } else {
+      # make sure random factors are not numeric, else, wrong "levels" will be returned
+      data[random_factors] <- lapply(data[random_factors], as.factor)
     }
   }
 

inst/WORDLIST

@@ -9,6 +9,7 @@ Biometrics
 Bundock
 CMD
 Coloured
+DCchoice
 DOI
 DirichletReg
 DirichletRegModel

tests/testthat/test-get_datagrid.R

@@ -394,3 +394,21 @@ test_that("get_datagrid - multiple weight variables", {
     tolerance = 1e-3
   )
 })
+
+
+test_that("get_datagrid - include_random works with numeric group factors", {
+  skip_if_not_installed("glmmTMB")
+  data(mtcars)
+  mtcars$vs <- as.factor(mtcars$vs)
+  model <- glmmTMB::glmmTMB(
+    mpg ~ vs + (1 | cyl),
+    data = mtcars
+  )
+  out <- get_datagrid(model, include_random = TRUE)
+  expect_identical(
+    out$cyl,
+    structure(c(1L, 1L, 2L, 2L, 3L, 3L), levels = c("4", "6", "8"), class = "factor")
+  )
+  out <- get_datagrid(model, include_random = FALSE)
+  expect_identical(out$cyl, c(NA, NA))
+})