Roll back changes done to the test 'simple corTestDS, some, with na, pearson', now replacing the input x, so the version with complete cases is c(1.0, 3.0, 7.0), instead of c(0.0, 3.0, 7.0). The latter resulted in a failure when testing for Infinite value on Apple Mx processors.

Author: Roberto Villegas-Diaz

tests/testthat/test-smk-corTestDS.R

@@ -182,7 +182,7 @@ test_that("simple corTestDS, some, pearson, without na, pearson", {
 
 context("corTestDS::smk::with na, pearson")
 test_that("simple corTestDS, some, with na, pearson", {
-    x   <- c(0.0, NA, 2.0, 3.0, NA, 5.0, NA, 7.0)
+    x   <- c(NA, 1.0, 2.0, 3.0, NA, 5.0, NA, 7.0)
     y   <- c(0.0, 1.0, NA, 3.0, 4.0, NA, NA, 7.0)
 
     res <- corTestDS("x", "y", "pearson", NULL, 0.95)
@@ -198,31 +198,7 @@ test_that("simple corTestDS, some, with na, pearson", {
 
     expect_equal(class(res$`Correlation test`$statistic), "numeric")
     expect_length(res$`Correlation test`$statistic, 1)
-    
-    # extract information from the results
-    t_stat <- res$`Correlation test`$statistic[[1]]
-    # degrees of freedom
-    df <- res$`Number of pairwise complete cases` - 2 
-    # re-calculate the value for r (correlation)
-    ## t_stat <- sqrt(df) * r / sqrt(1 - r^2) # Equation 1
-    ## calculate intermediate coefficient, derived from Equation 1
-    A <- t_stat / sqrt(df)
-    ## use the intermediate value to estimate the correlation value, r
-    r <- sqrt(A^2 / (1 + A^2))
-    
-    # if testing on Apple M1: due to numeric precision
-    if (getElement(Sys.info(), "sysname") == "Darwin" &&
-        getElement(R.version, "arch") == "aarch64") {
-      # t-stat should be a very large value, because r ~ 1, but not exactly 1
-      expect_false(is.infinite(t_stat))
-      # r ~ 1
-      expect_false(is.nan(r))
-    }
-    else { # other architectures
-      expect_true(is.infinite(t_stat))
-      # the estimated value for r should NaN
-      expect_true(is.nan(r))
-    }
+    expect_true(is.infinite(res$`Correlation test`$statistic[[1]]))
 
     expect_equal(class(res$`Correlation test`$parameter), "integer")
     expect_length(res$`Correlation test`$parameter, 1)