merge-recode

This is a design pattern for recoding variables using merges.

Description

• Intent. Provide a single, authoritative place for recoding variables.
• Scenario. You create different contrast variables during modeling fitting and want to put all of your recoding in a single place.
• Solution. Create functions that recode variables by merging observations with keys.

Recoders

Here is an example recoder function:

recode_feat_type <- function(frame) {
  feat_type_map <- data.frame(
    feat_type = c("visual", "nonvisual"),
    feat_c = c(-0.5, 0.5)
  )

  merge(frame, feat_type_map, all.x = TRUE)
}

Here's how to use it:

data <- data.frame(feat_type = c("visual", "visual", "nonvisual", "visual"))
data <- recode_feat_type(data)

Recoding multiple variables

The dplyr package exports a %>% function, which is very useful for chaining multiple recodings together:

library(dplyr)  # dplyr imports the `%>%` function from `magrittr`

clean_data <- dirty_data %>%
  recode_var1 %>%
  recode_var2 %>%
  recode_var3

Use case

Plotting model predictions on top of raw means. See "use-case.R".