Merge branch 'main' into strengejacke/issue31

DESCRIPTION

@@ -1,7 +1,7 @@
 Type: Package
 Package: bayestestR
 Title: Understand and Describe Bayesian Models and Posterior Distributions
-Version: 0.15.0.3
+Version: 0.15.0.4
 Authors@R:
     c(person(given = "Dominique",
              family = "Makowski",
@@ -95,7 +95,7 @@ Suggests:
     lavaan,
     lme4,
     logspline (>= 2.1.21),
-    marginaleffects (>= 0.24.0),
+    marginaleffects (>= 0.24.0.6),
     MASS,
     mclust,
     mediation,
@@ -128,4 +128,4 @@ Config/testthat/parallel: true
 Config/rcmdcheck/ignore-inconsequential-notes: true
 Config/Needs/website: easystats/easystatstemplate
 Config/Needs/check: stan-dev/cmdstanr
-Remotes: easystats/datawizard
+Remotes: easystats/datawizard, vincentarelbundock/marginaleffects

NEWS.md

@@ -1,3 +1,10 @@
+# bayestestR 0.15.0.4
+
+## Bug fixes
+
+* Fix to `emmeans` / `marginaleffects` / `data.frame(<rvar>)` methods when using multiple credible levels (#688).
+
+
 # bayestestR 0.15.0
 
 ## Changes

R/utils.R

@@ -230,7 +230,7 @@
   datagrid <- insight::get_datagrid(object)
   grid_names <- colnames(datagrid)
 
-  if (long) {
+  if (long || nrow(datagrid) < nrow(results)) {
     datagrid$Parameter <- unique(results$Parameter)
     results <- datawizard::data_merge(datagrid, results, by = "Parameter")
     results$Parameter <- NULL
@@ -270,7 +270,7 @@
   grid_names <- colnames(object)[!is_rvar]
   datagrid <- data.frame(object[, grid_names, drop = FALSE])
 
-  if (long) {
+  if (long || nrow(datagrid) < nrow(results)) {
     datagrid$Parameter <- unique(results$Parameter)
     results <- datawizard::data_merge(datagrid, results, by = "Parameter")
     results$Parameter <- NULL

tests/testthat/test-data.frame-with-rvar.R

@@ -45,6 +45,13 @@ test_that("data.frame w/ rvar_col descrive_posterior etc", {
   res <- eti(dfx, rvar_col = "my_rvar", ci = c(0.8, 0.95))
   res.ref <- eti(dfx$my_rvar, ci = c(0.8, 0.95))
   expect_true(all(c("mu", "sigma") %in% colnames(res)))
+  expect_identical(
+    as.data.frame(res[c("mu", "sigma")]),
+    data.frame(
+      mu = c(0, 0, 0.5, 0.5, 1, 1),
+      sigma = c(1, 1, 0.5, 0.5, 0.25, 0.25)
+    )
+  )
   expect_identical(nrow(format(res)), 3L)
   expect_identical(ncol(format(res)), 4L)
   expect_equal(res[setdiff(colnames(res), c("mu", "sigma"))],

tests/testthat/test-emmGrid.R

@@ -30,7 +30,16 @@ test_that("emmGrid hdi", {
   expect_equal(xhdi$CI_high, all_summ$upper.HPD, tolerance = 0.1)
 
   xhdi2 <- hdi(emc_, ci = 0.95)
-  expect_equal(xhdi$CI_low, xhdi2$CI_low)
+  expect_identical(xhdi$CI_low, xhdi2$CI_low)
+
+  xhdi3 <- hdi(all_, ci = c(0.9, 0.95))
+  expect_identical(
+    as.data.frame(xhdi3[1:2]),
+    data.frame(
+      group = c("1", "1", "2", "2", ".", "."),
+      contrast = c(".", ".", ".", ".", "group1 - group2", "group1 - group2"), stringsAsFactors = FALSE
+    )
+  )
 })
 
 test_that("emmGrid point_estimate", {
@@ -39,7 +48,7 @@ test_that("emmGrid point_estimate", {
   expect_equal(xpest$Median, all_summ$emmean, tolerance = 0.1)
 
   xpest2 <- point_estimate(emc_, centrality = "all", dispersion = TRUE)
-  expect_equal(xpest$Median, xpest2$Median)
+  expect_identical(xpest$Median, xpest2$Median)
 })
 
 
@@ -114,15 +123,15 @@ test_that("emmGrid rope", {
 # describe_posterior ------------------------------------------------------
 
 test_that("emmGrid describe_posterior", {
-  expect_equal(
+  expect_identical(
     describe_posterior(all_)$median,
     describe_posterior(emc_)$median
   )
 
   expect_identical(colnames(describe_posterior(all_))[1:2], c("group", "contrast"))
 
   skip_on_cran()
-  expect_equal(
+  expect_identical(
     describe_posterior(all_, bf_prior = model_p, test = "bf")$log_BF,
     describe_posterior(emc_, bf_prior = model_p, test = "bf")$log_BF
   )
@@ -167,9 +176,9 @@ test_that("emmGrid bayesfactor_restricted", {
   hyps <- c("`1` < `2`", "`1` < 0")
 
   xrbf <- bayesfactor_restricted(em_, prior = model_p, hypothesis = hyps)
-  expect_equal(length(xrbf$log_BF), 2)
-  expect_equal(length(xrbf$p_prior), 2)
-  expect_equal(length(xrbf$p_posterior), 2)
+  expect_length(xrbf$log_BF, 2)
+  expect_length(xrbf$p_prior, 2)
+  expect_length(xrbf$p_posterior, 2)
   expect_warning(bayesfactor_restricted(em_, hypothesis = hyps))
 
   xrbf2 <- bayesfactor_restricted(emc_, prior = model_p, hypothesis = hyps)
@@ -182,12 +191,12 @@ test_that("emmGrid si", {
 
   xrsi <- si(all_, prior = model_p, verbose = FALSE)
   expect_identical(colnames(xrsi)[1:2], c("group", "contrast"))
-  expect_equal(length(xrsi$CI_low), 3)
-  expect_equal(length(xrsi$CI_high), 3)
+  expect_length(xrsi$CI_low, 3)
+  expect_length(xrsi$CI_high, 3)
 
   xrsi2 <- si(emc_, prior = model_p, verbose = FALSE)
-  expect_equal(xrsi$CI_low, xrsi2$CI_low)
-  expect_equal(xrsi$CI_high, xrsi2$CI_high)
+  expect_identical(xrsi$CI_low, xrsi2$CI_low)
+  expect_identical(xrsi$CI_high, xrsi2$CI_high)
 })
 
 

tests/testthat/test-marginaleffects.R

@@ -26,11 +26,34 @@ withr::with_environment(
     )
 
     expect_true(all(c("term", "contrast") %in% colnames(results)))
-    expect_equal(results[setdiff(colnames(results), c("term", "contrast"))],
+    expect_equal(results[setdiff(colnames(results), c("term", "contrast", "am"))],
       results_draws[setdiff(colnames(results_draws), "Parameter")],
       ignore_attr = TRUE
     )
 
+    # multi ci levels
+    res <- hdi(mfx, ci = c(0.8, 0.9))
+    expect_identical(
+      as.data.frame(res[1:3]),
+      data.frame(
+        term = c(
+          "am", "am", "am", "am", "cyl", "cyl",
+          "cyl", "cyl", "cyl", "cyl", "cyl", "cyl",
+          "hp", "hp", "hp", "hp"
+        ),
+        contrast = c(
+          "1 - 0", "1 - 0", "1 - 0", "1 - 0",
+          "6 - 4", "6 - 4", "8 - 4", "8 - 4",
+          "6 - 4", "6 - 4", "8 - 4", "8 - 4",
+          "dY/dX", "dY/dX", "dY/dX", "dY/dX"
+        ),
+        am = c(
+          0, 0, 1, 1, 0, 0, 0, 0, 1, 1, 1, 1, 0, 0,
+          1, 1
+        ), stringsAsFactors = FALSE
+      )
+    )
+
     # estimate_density
     mfx <- marginaleffects::comparisons(mod,
       variables = "cyl",
@@ -67,16 +90,16 @@ withr::with_environment(
     outsi <- si(mfx, prior = mfxp, verbose = FALSE)
     outsiref <- si(mfx_samps, prior = mfxp_samps, verbose = FALSE)
 
-    expect_true(all(c("term", "contrast") %in% colnames(outsi)))
-    expect_equal(outsi[setdiff(colnames(outsi), c("term", "contrast"))],
+    expect_true(all(c("term", "contrast", "am") %in% colnames(outsi)))
+    expect_equal(outsi[setdiff(colnames(outsi), c("term", "contrast", "am"))],
       outsiref[setdiff(colnames(outsiref), "Parameter")],
       ignore_attr = TRUE
     )
 
     # bayesfactor_parameters
     bfp <- bayesfactor_parameters(mfx, prior = mfxp, verbose = FALSE)
     bfpref <- bayesfactor_parameters(mfx_samps, prior = mfxp_samps, verbose = FALSE)
-    expect_equal(bfp[setdiff(colnames(bfp), c("term", "contrast"))],
+    expect_equal(bfp[setdiff(colnames(bfp), c("term", "contrast", "am"))],
       bfpref[setdiff(colnames(bfpref), "Parameter")],
       ignore_attr = TRUE
     )

vignettes/bayes_factors.Rmd

@@ -121,7 +121,7 @@ Here we provide functions for computing Bayes factors in two different contexts:
 - **testing single parameters (coefficients) within a model** 
 - **comparing statistical models themselves**
 
-# Testing Models' Parameters with Bayes Factors {#bayesfactor_parameters}
+# 1. Testing Models' Parameters with Bayes Factors {#bayesfactor_parameters}
 
 A **Bayes factor for a single parameter** can be used to answer the question:
 
@@ -428,13 +428,8 @@ than the prior densities.
 
 In `bayestestR`, this can be achieved with the `si()` function:
 
-```{r}
-my_first_si <- si(
-  posterior = data.frame(group2 = posterior),
-  prior = data.frame(group2 = prior),
-  BF = 1,
-  verbose = FALSE
-)
+```{r, eval=FALSE}
+my_first_si <- si(model, BF = 1)
 
 print(my_first_si)
 ```
@@ -471,7 +466,7 @@ observing the data.
 decreased by observing the data. Testing against values outside this interval
 will produce a Bayes factor larger than $1/BF$ in support of the alternative.
 
-# Comparing Models using Bayes Factors {#bayesfactor_models}
+# 2. Comparing Models using Bayes Factors {#bayesfactor_models}
 
 Bayes factors can also be used to compare statistical **models**. In this
 statistical context, they answer the following question:
@@ -738,7 +733,7 @@ $$
 
 **NOTE**: See the *Specifying Correct Priors for Factors with More Than 2 Levels* appendix below.
 
-# Bayesian Model Averaging
+# 3. Bayesian Model Averaging
 
 In the previous section, we discussed the direct comparison of two models to
 determine if an effect is supported by the data. However, in many cases there
@@ -792,7 +787,9 @@ bayesfactor_inclusion(comparison, match_models = TRUE)
 `bayesfactor_inclusion()` is meant to provide Bayes Factors per predictor,
 similar to JASP's *Effects* option. 
 
-Let's compare the two:
+Let's compare the two. 
+Note that for this comparison we will use the `{BayesFactor}` package, which is what _JASP_ uses under the hood.
+(Note that this package used different model-parameterization and different default prior-specifications compared to _Stan_-based packages.)
 
 1. **Across all models**: