Split response names with dots in contrasts of multivariate brms models

DESCRIPTION

@@ -1,7 +1,7 @@
 Type: Package
 Package: modelbased
 Title: Estimation of Model-Based Predictions, Contrasts and Means
-Version: 0.17.0.18
+Version: 0.17.0.19
 Authors@R:
     c(person(given = "Dominique",
              family = "Makowski",

NEWS.md

@@ -170,6 +170,12 @@
   `estimate_contrasts()` now uses the names of the response variables, not
   their observed values, to find the response in the contrast labels.
 
+* `estimate_contrasts()` now returns separate `Response1` and `Response2`
+  columns for multivariate *brms* models with dots in the response names, for
+  example `mvbind(Sepal.Length, Petal.Length) ~ Species`. Before, the response
+  names stayed in the `Level1` and `Level2` columns, because *brms* removes the
+  dots from the response names.
+
 * The `grid` argument of `visualisation_recipe()` and `plot()` now changes the
   facet grid of plots with four or more `by` variables. Before, these functions
   applied the `facet` argument to the facet grid instead. If `facet` was not

R/format.R

@@ -646,7 +646,15 @@ format.marginaleffects_contrasts <- function(
     response_levels <- unique(as.character(response))
   }
   if (is_multivariate) {
-    response_levels <- unique(c(response_levels, insight::find_response(model)))
+    # brms drops dots from response names (e.g., "Sepal.Length" becomes
+    # "SepalLength"), and uses these names in the contrast labels. The names
+    # of `find_response()` hold these brms names
+    model_response <- insight::find_response(model)
+    response_levels <- unique(c(
+      response_levels,
+      model_response,
+      names(model_response)
+    ))
   }
   if (!length(response_levels)) {
     return(params)

tests/testthat/test-brms.R

@@ -117,6 +117,27 @@ test_that("estimate_contrasts - brms, multivariate, many observations", {
 })
 
 
+test_that("estimate_contrasts - brms, multivariate, response names with dots", {
+  # brms drops the dot from response names ("Sepal.Length" -> "SepalLength"),
+  # and the contrast labels use the brms names
+  m <- suppressWarnings(brms::brm(
+    brms::mvbind(Sepal.Length, Petal.Length) ~ Species,
+    data = iris,
+    refresh = 0,
+    iter = 500,
+    chains = 1,
+    seed = 123
+  ))
+
+  out <- estimate_contrasts(m, "Species")
+  expect_identical(nrow(out), 15L)
+  expect_true(all(c("Response1", "Response2") %in% colnames(out)))
+  expect_identical(levels(out$Response1), c("SepalLength", "PetalLength"))
+  expect_true(all(as.character(out$Level1) %in% levels(iris$Species)))
+  expect_true(all(as.character(out$Level2) %in% levels(iris$Species)))
+})
+
+
 test_that("estimate_means - brms, Wiener", {
   skip_if_not_installed("curl")
   skip_if_offline()