trichotomize.R

# Sometimes, you will find useful to discretize a variable.
# An option to do that is creating three equal sized groups
# this is, trichotomization. 
# Reference http://www.stat.columbia.edu/~gelman/research/published/thirds5.pdf

library(Hmisc) 

# We use cut2 to create three equal groups
# then, by taking the number with as.numeric
# we will have a column filled with the group membership
# as 1, 2 and 3.
mtcars$wt_q <- as.numeric(cut2(mtcars$wt, g = 3))
# Substracting 2 we get the group membership coded 
# as -1, 0 and 1. Then, we convert it in a factor
mtcars$wt_q <- as.factor(mtcars$wt_q - 2)
# We set the central group as reference level 
mtcars$wt_q = relevel(mtcars$wt_q, ref="0")