bug in winsorized Pearson's correlation

[IndrajeetPatil]

There are two issues here:

• the correlation coefficient should be the same across these two contexts, but they are not
• setting p_adjust = "none" doesn't do anything

library(WRS2)
library(tidyverse)
library(correlation)

correlation(select(msleep, bodywt, sleep_rem), winsorize = 0.2)
#> # Correlation table (pearson-method)
#> 
#> Parameter1 | Parameter2 |     r |         95% CI | t(59) |       p
#> ------------------------------------------------------------------
#> bodywt     |  sleep_rem | -0.37 | [-0.57, -0.14] | -3.10 | 0.003**
#> 
#> p-value adjustment method: Holm (1979)
#> Observations: 61

correlation(keep(.x = msleep, .p = is_bare_numeric), winsorize = 0.2, p_adjust = "none") %>%
  filter(Parameter1 == "sleep_rem", Parameter2 == "bodywt")
#> # Correlation table (pearson-method)
#> 
#> Parameter1 | Parameter2 |     r |        95% CI | t(28) |     p
#> ---------------------------------------------------------------
#> sleep_rem  |     bodywt | -0.35 | [-0.63, 0.01] | -1.97 | 0.059
#> 
#> p-value adjustment method: none
#> Observations: 30

^{Created on 2021-03-19 by the reprex package (v1.0.0)}

[IndrajeetPatil]

Hmm, seems to be due to how NA's are being omitted. The number of observations for the first test is 61, while 30 for the second one.

correlation/R/cor_test.R

Line 163 in 9766bcd

data <- stats::na.omit(data)

[IndrajeetPatil]

With the PR:

library(WRS2)
library(tidyverse)
library(correlation)

correlation(select(msleep, bodywt, sleep_rem), winsorize = 0.2)
#> # Correlation Matrix (pearson-method)
#> 
#> Parameter1 | Parameter2 |     r |         95% CI | t(59) |       p
#> ------------------------------------------------------------------
#> bodywt     |  sleep_rem | -0.37 | [-0.57, -0.14] | -3.10 | < .01**
#> 
#> p-value adjustment method: Holm (1979)
#> Observations: 61

correlation(keep(.x = msleep, .p = is_bare_numeric), winsorize = 0.2, p_adjust = "none") %>%
  filter(Parameter1 == "sleep_rem", Parameter2 == "bodywt")
#> # Correlation Matrix (pearson-method)
#> 
#> Parameter1 | Parameter2 |     r |         95% CI | t(59) |       p
#> ------------------------------------------------------------------
#> sleep_rem  |     bodywt | -0.37 | [-0.57, -0.14] | -3.10 | < .01**
#> 
#> p-value adjustment method: none
#> Observations: 61

^{Created on 2021-03-19 by the reprex package (v1.0.0)}