Pairwise Correlation Tests Across All Numeric Columns

Runs cor.test on every pair of numeric columns in a data frame and returns one row per pair.

Usage

corr_test_all(
  data,
  method = "pearson",
  use = c("pairwise.complete.obs", "complete.obs"),
  columns = c("compact", "tidy"),
  sort = c("p_value", "estimate", "abs_estimate", "none"),
  ...
)

Arguments

data

A data frame or tibble. Non-numeric columns are ignored.

method

Correlation method passed to cor.test: "pearson" (default), "kendall", or "spearman".

use

How missing values are handled. "pairwise.complete.obs" (default) tests each pair on the rows where both columns are observed, which is what cor.test() does natively. "complete.obs" first drops every row with a missing value in any numeric column, so all pairs are tested on the same rows.

columns

Output shape. "compact" (default) returns var1, var2, r, p_value. "tidy" returns var1, var2, then every column broom::tidy() reports for the test, with estimate renamed cor: cor, statistic, p.value, parameter, conf.low, conf.high, method, alternative. parameter and the confidence limits are only reported by methods that compute them (Pearson).

sort

Row order. "p_value" (default) is ascending p-value; "estimate" is descending correlation; "abs_estimate" is descending absolute correlation; "none" keeps pair order. Ties keep pair order, and NAs sort last.

...

Further arguments passed to cor.test, such as alternative, conf.level, or exact.

Value

A tibble with one row per pair of numeric columns.

Details

Pairs are enumerated in column order (via combn), so var1 is always the column that appears first in data. A pair whose test cannot be computed (e.g. fewer than three complete observations) is kept with NA results rather than aborting the whole run. Invalid arguments (an unknown method, alternative, a bad conf.level, ...) are errors, not NA rows.

Examples

corr_test_all(iris[, 1:4])
#> # A tibble: 6 × 4
#>   var1         var2              r  p_value
#>   <chr>        <chr>         <dbl>    <dbl>
#> 1 Petal.Length Petal.Width   0.963 4.68e-86
#> 2 Sepal.Length Petal.Length  0.872 1.04e-47
#> 3 Sepal.Length Petal.Width   0.818 2.33e-37
#> 4 Sepal.Width  Petal.Length -0.428 4.51e- 8
#> 5 Sepal.Width  Petal.Width  -0.366 4.07e- 6
#> 6 Sepal.Length Sepal.Width  -0.118 1.52e- 1

# Every cor.test() statistic, strongest positive correlation first
corr_test_all(mtcars[, c("mpg", "hp", "wt", "qsec")], columns = "tidy", sort = "estimate")
#> # A tibble: 6 × 10
#>   var1  var2     cor statistic  p.value parameter conf.low conf.high method     
#>   <chr> <chr>  <dbl>     <dbl>    <dbl>     <int>    <dbl>     <dbl> <chr>      
#> 1 hp    wt     0.659     4.80  4.15e- 5        30   0.403      0.819 Pearson's …
#> 2 mpg   qsec   0.419     2.53  1.71e- 2        30   0.0820     0.670 Pearson's …
#> 3 wt    qsec  -0.175    -0.972 3.39e- 1        30  -0.493      0.185 Pearson's …
#> 4 hp    qsec  -0.708    -5.49  5.77e- 6        30  -0.848     -0.477 Pearson's …
#> 5 mpg   hp    -0.776    -6.74  1.79e- 7        30  -0.885     -0.586 Pearson's …
#> 6 mpg   wt    -0.868    -9.56  1.29e-10        30  -0.934     -0.744 Pearson's …
#> # ℹ 1 more variable: alternative <chr>