polychoric correlation method yields error

DESCRIPTION

@@ -1,7 +1,7 @@
 Type: Package
 Package: correlation
 Title: Methods for Correlation Analysis
-Version: 0.8.7.1
+Version: 0.8.7.2
 Authors@R:
     c(person(given = "Dominique",
              family = "Makowski",

NEWS.md

@@ -2,6 +2,8 @@
 
 - `correlation()` gains a `missing=` argument, similar to `stats::cor(use=)`, for controlling how missing data is handled.
 
+- `correlation()` converts numeric input variables automatically into factors when `method = "polychoric"`.
+
 # correlation 0.8.7
 
 - The `format()` method for objects of class `easycormatrix` gets a `zap_small`

R/cor_test.R

@@ -1,7 +1,7 @@
 #' Correlation test
 #'
 #' This function performs a correlation test between two variables.
 #' You can easily visualize the result using [`plot()`][visualisation_recipe.easycormatrix()] (see examples [**here**](https://easystats.github.io/correlation/reference/visualisation_recipe.easycormatrix.html#ref-examples)).
 #'
 #' @param data A data frame.
 #' @param x,y Names of two variables present in the data.
@@ -118,7 +118,7 @@
 #' }
 #' }
 #' @export
 cor_test <- function(data,
                      x,
                      y,
                      method = "pearson",
@@ -137,7 +137,7 @@
                      ...) {
   # valid matrix checks
   if (!all(x %in% names(data)) || !all(y %in% names(data))) {
     insight::format_error("The names you entered for x and y are not available in the dataset. Make sure there are no typos!")
   }
 
   if (ci == "default") ci <- 0.95
@@ -148,6 +148,13 @@
     data[c(x, y)] <- datawizard::to_numeric(data[c(x, y)], dummy_factors = FALSE)
   }
 
+  # However, for poly, we need factors!
+  if (method %in% c("poly", "polychoric") && all(vapply(data[c(x, y)], is.numeric, FUN.VALUE = TRUE))) {
+    # convert all input to factors, but only if all input currently is numeric
+    # we allow mix of numeric and factors
+    data[c(x, y)] <- datawizard::to_factor(data[c(x, y)])
+  }
+
   # Partial
   if (!isFALSE(partial)) {
     # partial
@@ -237,19 +244,19 @@
 
     # Bayesian
   } else if (method %in% c("tetra", "tetrachoric")) {
     insight::format_error("Tetrachoric Bayesian correlations are not supported yet. Get in touch if you want to contribute.")
   } else if (method %in% c("poly", "polychoric")) {
     insight::format_error("Polychoric Bayesian correlations are not supported yet. Get in touch if you want to contribute.")
   } else if (method %in% c("biserial", "pointbiserial", "point-biserial")) {
     insight::format_error("Biserial Bayesian correlations are not supported yet. Get in touch if you want to contribute.")
   } else if (method == "biweight") {
     insight::format_error("Biweight Bayesian correlations are not supported yet. Get in touch if you want to contribute.")
   } else if (method == "distance") {
     insight::format_error("Bayesian distance correlations are not supported yet. Get in touch if you want to contribute.")
   } else if (method %in% c("percentage", "percentage_bend", "percentagebend", "pb")) {
     insight::format_error("Bayesian Percentage Bend correlations are not supported yet. Get in touch if you want to contribute.")
   } else if (method %in% c("blomqvist", "median", "medial")) {
     insight::format_error("Bayesian Blomqvist correlations are not supported yet. Check-out the BBcor package (https://github.com/donaldRwilliams/BBcor).")
   } else if (method == "hoeffding") {
     insight::format_error("Bayesian Hoeffding's correlations are not supported yet. Check-out the BBcor package (https://github.com/donaldRwilliams/BBcor).")
   } else if (method == "gamma") {

R/correlation.R

@@ -551,7 +551,15 @@ correlation <- function(data,
     include_factors <- TRUE
   }
 
-  if (method == "polychoric") multilevel <- TRUE
+  # definitely need factors for polychoric
+  if (method == "polychoric") {
+    multilevel <- TRUE
+    # convert all input to factors, but only if all input currently is numeric
+    # we allow mix of numeric and factors
+    if (all(vapply(data, is.numeric, FUN.VALUE = TRUE))) {
+      data <- datawizard::to_factor(data)
+    }
+  }
 
   # Clean data and get combinations -------------
 

tests/testthat/test-cor_test.R

@@ -88,6 +88,23 @@ test_that("cor_test tetrachoric", {
   out <- cor_test(data, "Sepal.Width", "Petal.Width_binary", method = "biserial")
   expect_equal(out$rho, -0.403, tolerance = 0.01)
   out_psych <- psych::biserial(data[["Sepal.Width"]], data[["Petal.Width_binary"]])[1]
+
+  set.seed(123)
+  n <- 100
+  k <- 5
+  d <- data.frame(
+    x1 = sample.int(k, n, TRUE),
+    x2 = sample.int(k, n, TRUE),
+    x3 = sample.int(k, n, TRUE),
+    x4 = sample.int(k, n, TRUE)
+  )
+  expect_true(all(vapply(d, is.numeric, logical(1))))
+  out <- correlation(d, method = "polychoric")
+  expect_equal(
+    out$rho,
+    c(0.07729, -0.02453, -0.13999, 0.06508, -0.17158, 0.17863),
+    tolerance = 1e-3
+  )
 })