Skip to main content

Numeric coding of a variable for a correlation

Description

  • logical -> 0/1;

  • binary factor or character -> 0/1, the upper level coded 1 (so r > 0 means "higher in the upper level"); binary numeric -> 0/1 (largest value = 1);

  • ordered factor with > 2 levels -> 0, 1, 2, … (Pearson then assumes equal spacing between levels: warning);

  • unordered factor with > 2 levels -> NA with a warning (no meaningful correlation).

Usage

as_cor_numeric(x, name = "x", meth = "spearman")

Arguments

x vector to code
name name used in the warnings
meth correlation method (“pearson” triggers the equal-spacing warning)

Details

Recoding a binary variable to 0/1 is a positive linear transformation: it changes neither r, nor r_s, nor tau, nor their p-values.

Value

numeric vector

Examples

library("corrkit")

as_cor_numeric(factor(c("low", "high", "low"), levels = c("low", "high")))
[1] 0 1 0
Back to top