move testthat to 3rd ed. and update vignette with info on interactions

DESCRIPTION

@@ -1,39 +1,40 @@
-Package: multilevelTools
-Title: Multilevel and Mixed Effects Model Diagnostics and Effect Sizes
-Version: 0.1.3
-Date: 2023-02-10
-Authors@R: 
-    person(given = "Joshua F.",
-           family = "Wiley",
-           role = c("aut", "cre"),
-           email = "jwiley.psych@gmail.com",
-           comment = c(ORCID = "0000-0002-0271-6702"))	   
-URL: http://joshuawiley.com/multilevelTools, https://github.com/JWiley/multilevelTools
-BugReports: https://github.com/JWiley/multilevelTools/issues	   
-Description: Effect sizes, diagnostics and performance metrics for
-  multilevel and mixed effects models.
-  Includes marginal and conditional 'R2' estimates for linear mixed effects models
-  based on Johnson (2014) <doi:10.1111/2041-210X.12225>.
-License: GPL (>= 3)
-Depends:
-    R (>= 3.5.0)
-Imports:
-    lme4,
-    lmerTest,
-    data.table (>= 1.12.0),
-    nlme,
-    extraoperators (>= 0.1.1),
-    JWileymisc (>= 1.1.0),
-    cowplot,
-    ggplot2,
-    lavaan,
-    zoo
-Suggests: 
-    testthat (>= 2.1.0),
-    covr,
-    knitr,
-    rmarkdown
-Encoding: UTF-8
-LazyData: true
-VignetteBuilder: knitr
-RoxygenNote: 7.1.0
+Package: multilevelTools
+Title: Multilevel and Mixed Effects Model Diagnostics and Effect Sizes
+Version: 0.1.3
+Date: 2023-02-10
+Authors@R: 
+    person(given = "Joshua F.",
+           family = "Wiley",
+           role = c("aut", "cre"),
+           email = "jwiley.psych@gmail.com",
+           comment = c(ORCID = "0000-0002-0271-6702"))	   
+URL: https://joshuawiley.com/multilevelTools, https://github.com/JWiley/multilevelTools
+BugReports: https://github.com/JWiley/multilevelTools/issues	   
+Description: Effect sizes, diagnostics and performance metrics for
+  multilevel and mixed effects models.
+  Includes marginal and conditional 'R2' estimates for linear mixed effects models
+  based on Johnson (2014) <doi:10.1111/2041-210X.12225>.
+License: GPL (>= 3)
+Depends:
+    R (>= 4.0.0)
+Imports:
+    lme4,
+    lmerTest,
+    data.table (>= 1.14.6),
+    nlme,
+    extraoperators (>= 0.1.1),
+    JWileymisc (>= 1.4.0),
+    cowplot,
+    ggplot2,
+    lavaan,
+    zoo
+Suggests: 
+    testthat (>= 3.1.0),
+    covr,
+    knitr,
+    rmarkdown
+Encoding: UTF-8
+LazyData: true
+Config/testthat/edition: 3
+VignetteBuilder: knitr
+RoxygenNote: 7.2.3

NEWS.md

@@ -1,3 +1,18 @@
+# multilevelTools 0.1.3
+
+## Bug Fixes
+
+* `modelTest()` no longer fails for models with a continuous x categorical interaction.
+  Estimates for dropping the "simple" effect of the continuous variable are still 
+  not calculable, but the rest of the calculations are still performed and that line
+  is simply set to NA.
+
+## Changes
+
+* moved to testthat 3rd edition
+* moved CI to GitHub actions
+* use preferably for package website
+
 # multilevelTools 0.1.2
 
 ## New Features

R/models.R

@@ -706,14 +706,21 @@ modelTest.merMod <- function(object, method = c("Wald", "profile", "boot"), cont
     objreduced <- testm[[i]]
     v <- labs.levels$Terms[[i]]
     nparreduced <- attr(logLik(objreduced), "df")
-
-    if (!isTRUE(inherits(objreduced, "merMod")) | nparreduced >= nparfull) {
+
+    if (nparreduced >= nparfull) {
+      msg <- sprintf("The full and reduced model had %d and %d parameters, respectively.\nThe reduced model should have fewer parameters.\nThis usually happens when there are interactions with categorical variables.\nFor an explanation, see:\nhttps://joshuawiley.com/multilevelTools/articles/lmer-vignette.html", nparfull, nparreduced)
+      message(msg)
       tmp <- copy(NAtemplate)
     } else {
-      tmp <- modelCompare(object, objreduced)$Comparison[3]
+
+      if (!isTRUE(inherits(objreduced, "merMod"))) {
+        tmp <- copy(NAtemplate)
+      } else {
+        tmp <- modelCompare(object, objreduced)$Comparison[3]
+      }
     }
     tmp$Model <- v
-
+    
     setnames(tmp, old = "Model", new = "Variable")
     return(tmp)
   }))

README.md

@@ -2,12 +2,12 @@ multilevelTools
 ===============
 
 <!-- badges: start -->
-[![Build
-Status](https://travis-ci.com/JWiley/multilevelTools.svg?branch=master)](https://travis-ci.com/JWiley/multilevelTools)
-[![Codecov test coverage](https://codecov.io/gh/JWiley/multilevelTools/branch/master/graph/badge.svg)](https://codecov.io/gh/JWiley/multilevelTools?branch=master)
+[![CRAN_status](https://www.r-pkg.org/badges/version/multilevelTools)](https://cran.r-project.org/package=multilevelTools)
+[![R-CMD-check](https://github.com/JWiley/multilevelTools/workflows/R-CMD-check/badge.svg)](https://github.com/JWiley/multilevelTools/actions)
+[![codecov](https://codecov.io/gh/JWiley/multilevelTools/branch/main/graph/badge.svg?token=HK1Ns92mfB)](https://codecov.io/gh/JWiley/multilevelTools)
+[![lifecycle](https://img.shields.io/badge/lifecycle-stable-brightgreen.svg)](https://lifecycle.r-lib.org/articles/stages.html#stable)
 <!-- badges: end -->
 
-
 Effect sizes, diagnostics and performance metrics for
 multilevel and mixed effects models.
 Includes marginal and conditional *R2* estimates for linear mixed

tests/testthat/test-descriptives.R

@@ -5,13 +5,13 @@ test_that("meanDecompose works with one grouping factor", {
   expect_length(md, 2L)
 
   ## residuals are same as original data
-  expect_equivalent(nrow(md[[length(md)]]), nrow(mtcars))
+  expect_equal(nrow(md[[length(md)]]), nrow(mtcars))
 
   ## cyl means are equal to unique cyl
-  expect_equivalent(nrow(md[[1]]), length(unique(mtcars$cyl)))
+  expect_equal(nrow(md[[1]]), length(unique(mtcars$cyl)))
 
   ## residuals have mean about 0
-  expect_equivalent(mean(md[[length(md)]]$X, na.rm = TRUE),
+  expect_equal(mean(md[[length(md)]]$X, na.rm = TRUE),
                     0, tolerance = .05)
 })
 
@@ -22,61 +22,61 @@ test_that("meanDecompose works with two grouping factors", {
   expect_length(md, 3L)
 
   ## residuals are same as original data
-  expect_equivalent(nrow(md[[length(md)]]), nrow(mtcars))
+  expect_identical(nrow(md[[length(md)]]), nrow(mtcars))
 
   ## vs means are equal to unique vs
-  expect_equivalent(nrow(md[[1]]), length(unique(mtcars$vs)))
+  expect_identical(nrow(md[[1]]), length(unique(mtcars$vs)))
 
   ## vs by cyl means are equal to unique vs x cyl
-  expect_equivalent(nrow(md[[2]]), length(unique(paste0(mtcars$vs, mtcars$cyl))))
+  expect_identical(nrow(md[[2]]), length(unique(paste0(mtcars$vs, mtcars$cyl))))
 
   ## residuals have mean about 0
-  expect_equivalent(mean(md[[length(md)]]$X, na.rm = TRUE),
+  expect_equal(mean(md[[length(md)]]$X, na.rm = TRUE),
                     0, tolerance = .05)
 })
 
 
 test_that("iccMixed works with one grouping factor", {
   icc <- iccMixed("mpg", "cyl", mtcars)
 
-  expect_equivalent(nrow(icc), 2)
-  expect_equivalent(
+  expect_identical(nrow(icc), 2L)
+  expect_equal(
     sum(icc$ICC, na.rm = TRUE),
     1, tolerance = .02)
 })
 
 test_that("iccMixed works with one grouping factor and data.table", {
   icc <- iccMixed("mpg", "cyl", data.table::as.data.table(mtcars))
 
-  expect_equivalent(nrow(icc), 2)
-  expect_equivalent(
+  expect_identical(nrow(icc), 2L)
+  expect_equal(
     sum(icc$ICC, na.rm = TRUE),
     1, tolerance = .02)
 })
 
 test_that("iccMixed works with two grouping factors", {
   icc <- iccMixed("mpg", c("cyl", "am"), mtcars)
 
-  expect_equivalent(nrow(icc), 3)
-  expect_equivalent(
+  expect_identical(nrow(icc), 3L)
+  expect_equal(
     sum(icc$ICC, na.rm = TRUE),
     1, tolerance = .02)
 })
 
 test_that("iccMixed works with two grouping factors and data.table", {
   icc <- iccMixed("mpg", c("cyl", "am"), data.table::as.data.table(mtcars))
 
-  expect_equivalent(nrow(icc), 3)
-  expect_equivalent(
+  expect_identical(nrow(icc), 3L)
+  expect_equal(
     sum(icc$ICC, na.rm = TRUE),
     1, tolerance = .02)
 })
 
 test_that("iccMixed works with one grouping factor and binomial family", {
   icc <- iccMixed("am", "cyl", mtcars)
 
-  expect_equivalent(nrow(icc), 2)
-  expect_equivalent(
+  expect_identical(nrow(icc), 2L)
+  expect_equal(
     sum(icc$ICC, na.rm = TRUE),
     1, tolerance = .02)
 })
@@ -85,35 +85,35 @@ test_that("iccMixed works with one grouping factor and binomial family", {
 test_that("nEffective works with data & gaussian (default) & data.frame", {
   nk <- nEffective(dv = "mpg", id = "cyl", data = mtcars)
 
-  expect_equivalent(nrow(nk), 3)
+  expect_identical(nrow(nk), 3L)
   expect_true(all(nk$N[1:2] <= nk$N[3]))
 })
 
 test_that("nEffective works with data & binomial & data.frame", {
   nk <- nEffective(dv = "am", id = "cyl", data = mtcars, family = "binomial")
 
-  expect_equivalent(nrow(nk), 3)
+  expect_identical(nrow(nk), 3L)
   expect_true(all(nk$N[1:2] <= nk$N[3]))
 })
 
 test_that("nEffective works with data & gaussian (default) & data.table", {
   nk <- nEffective(dv = "mpg", id = "cyl", data = data.table::as.data.table(mtcars))
 
-  expect_equivalent(nrow(nk), 3)
+  expect_identical(nrow(nk), 3L)
   expect_true(all(nk$N[1:2] <= nk$N[3]))
 })
 
 test_that("nEffective works with inputs & gaussian (default)", {
   nk <- nEffective(n = 60, k = 10, icc = .6)
 
-  expect_equivalent(nrow(nk), 3)
+  expect_identical(nrow(nk), 3L)
   expect_true(all(nk$N[1:2] <= nk$N[3]))
 })
 
 test_that("nEffective works with inputs & binomial", {
   nk <- nEffective(n = 60, k = 10, icc = .6, family = "binomial")
 
-  expect_equivalent(nrow(nk), 3)
+  expect_identical(nrow(nk), 3L)
   expect_true(all(nk$N[1:2] <= nk$N[3]))
 })
 
@@ -122,9 +122,9 @@ test_that("meanDeviations works", {
 
   expect_length(md, 2)
 
-  expect_equivalent(md[[1]], 5.5, tolerance = .1)
+  expect_equal(md[[1]], 5.5, tolerance = .1)
 
-  expect_equivalent(mean(md[[2]], na.rm = TRUE),
+  expect_equal(mean(md[[2]], na.rm = TRUE),
     0, tolerance = .1)
 })
 
@@ -141,7 +141,7 @@ test_that("acfByID works", {
   ac <- acfByID("x", "time", "id", data = dat, lag.max = 10L)
 
   ## should have 10 + 1 rows for lag 0 to 10
-  expect_equivalent(nrow(ac), 10 + 1)
+  expect_equal(nrow(ac), 10 + 1)
   ## correlations so expect between +/- 1
   ## add .1 for tolerance
   expect_true(all(ac$AutoCorrelation >= -1.1 & ac$AutoCorrelation <= 1.1))
@@ -150,7 +150,7 @@ test_that("acfByID works", {
   ac <- acfByID("x", "time", data = dat, lag.max = 10L)
 
   ## should have 10 + 1 rows for lag 0 to 10
-  expect_equivalent(nrow(ac), 10 + 1)
+  expect_equal(nrow(ac), 10 + 1)
   ## correlations so expect between +/- 1
   ## add .1 for tolerance
   expect_true(all(ac$AutoCorrelation >= -1.1 & ac$AutoCorrelation <= 1.1))
@@ -160,7 +160,7 @@ test_that("acfByID works", {
   ac <- acfByID("x", "time", "id", data = as.data.frame(dat), lag.max = 10L)
 
   ## should have 10 + 1 rows for lag 0 to 10
-  expect_equivalent(nrow(ac), 10 + 1)
+  expect_equal(nrow(ac), 10 + 1)
   ## correlations so expect between +/- 1
   ## add .1 for tolerance
   expect_true(all(ac$AutoCorrelation >= -1.1 & ac$AutoCorrelation <= 1.1))