In Ajfrick/ajfhelpR: General Assistance for EDA and table generation

library(tidyverse)
library(ajfhelpR)
knitr::opts_chunk$set(echo = TRUE)

One of the first "oops" mistakes I had in R had to do with T | NA == NA instead of false. While there are times when you don't want to coerce that to false, I decided to create a new set of boolean operators for this case.

First we'll look at the new equal to operator %==% by creating a data frame of boolean combinations

#Create data frame/tibble with combination of all bools

x = c(T,F,NA)

bools = as.tibble(expand.grid(x,x))

equals = bools %>% 
  mutate(`==` = (Var1 == Var2),
         `%==%` = (Var1 %==% Var2))

Below you can see the difference in the output for the original == and the %==% extension.

equals

Here we'll create another tibble containing the 'venn diagram' boolean operators

1. | vs %or%
2. & vs %&%

Venns = bools %>% 
  mutate(`|` = (Var1 | Var2),
         `%or%` = (Var1 %or% Var2),
         `&` = (Var1 & Var2),
         `%&%` = (Var1 %&% Var2))

Venns

The equality operator also allows comparison of a single value vs a vector, it can be used for subsetting operations.

x[(condition)] returns all values where (condition) is not false. NA is not interpreted as false by R, so any rows with missing condition will be also be returned. To ensure your subset only returns those that are true, we can use the %==% operator to coerce those to false.

set.seed(12345)

y = 'blue'

colors = c('blue', 'red', 'yellow', NA)

colorsamp = sample(colors, size = 10, prob = c(0.4,0.2,0.1,0.3),
                   replace = T)


y  ==  colorsamp
y %==% colorsamp

colorsamp[colorsamp == y]

colorsamp[colorsamp %==% y]

The %==% operator is especially useful for ifelse statements. Let's say we change all y to 0 if x is not equal to 'a'

set.seed(12356)

tib = tibble(
  x = sample(c("a", "b", NA), size = 10, replace = T),
  y = sample(1:10, size = 10, replace = T)
)

tib

We see there are 3 rows with NA, and depending on how we wish to view the data in context, we could consider these to be 1. "not equal to a" or 2. "unknown."

tib = tib %>% 
  mutate(`y %==%` = ifelse(x %==% 'a', x, 0),
         `y ==`   = ifelse(x  ==  'a', x, 0))

tib

The == operator perceives x == NA as missing and outputs a value of missing for the new column. The %==% opertor coerces those missing booleans to F, and outputs the value of 0. Either of these approaches can be correct based on the context of the individual problem. But I've found in data that I work with, %==% tends to be more useful.

Ajfrick/ajfhelpR documentation built on June 30, 2023, 12:56 a.m.