Recode variables

TODO

• link to factors

Install required packages

dplyr, forcats

wants <- c("dplyr", "forcats")
has   <- wants %in% rownames(installed.packages())
if(any(!has)) install.packages(wants[!has])

Recode numerical or character variables

Using index vectors

myColors <- c("red", "purple", "blue", "blue", "orange", "red", "orange")
farben   <- character(length(myColors))
farben[myColors == "red"]    <- "rot"
farben[myColors == "purple"] <- "violett"
farben[myColors == "blue"]   <- "blau"
farben[myColors == "orange"] <- "orange"
farben

[1] "rot"     "violett" "blau"    "blau"    "orange"  "rot"     "orange" 

replace(c(1, 2, 3, 4, 5), list=c(2, 4), values=c(200, 400))

[1]   1 200   3 400   5

Using recode() from package dplyr

library(dplyr)
recode(myColors, red="rot", blue="blau", purple="violett")

[1] "rot"     "violett" "blau"    "blau"    "orange"  "rot"     "orange" 

recode(myColors, red="basic", blue="basic", .default="complex")

[1] "basic"   "complex" "basic"   "basic"   "complex" "basic"   "complex"

Using ifelse()

orgVec <- c(5, 9, 11, 8, 9, 3, 1, 13, 9, 12, 5, 12, 6, 3, 17, 5, 8, 7)
cutoff <- 10
(reVec <- ifelse(orgVec <= cutoff, orgVec, cutoff))

 [1]  5  9 10  8  9  3  1 10  9 10  5 10  6  3 10  5  8  7

targetSet <- c("A", "B", "C", "D", "E", "F", "G", "H", "I", "J", "K")
response  <- c("Z", "E", "O", "W", "H", "C", "I", "G", "A", "O", "B")
(respRec  <- ifelse(response %in% targetSet, response, "other"))

 [1] "other" "E"     "other" "other" "H"     "C"     "I"     "G"     "A"    
[10] "other" "B"    

Cut continuous variables into categorical variables

Free recoding of value ranges into categories

Two intervals

set.seed(123)
IQ <- rnorm(20, mean=100, sd=15)
ifelse(IQ >= 100, "hi", "lo")

 [1] "lo" "lo" "hi" "hi" "hi" "hi" "hi" "lo" "lo" "lo" "hi" "hi" "hi" "hi" "lo"
[16] "hi" "hi" "lo" "hi" "lo"

Multiple intervals

library(dplyr)
case_when((IQ <= 100)              ~ 1,
          (IQ > 100) & (IQ <= 115) ~ 2,
          (IQ > 115)               ~ 3)

 [1] 1 1 3 2 2 3 2 1 1 1 3 2 2 2 1 3 2 1 2 1

Turn ordered value ranges into factor levels using cut()

IQfac <- cut(IQ, breaks=c(0, 85, 115, Inf), labels=c("lo", "mid", "hi"))
head(IQfac)

[1] mid mid hi  mid mid hi 
Levels: lo mid hi

medSplit <- cut(IQ, breaks=c(-Inf, median(IQ), Inf))
summary(medSplit)

(-Inf,102] (102, Inf] 
        10         10 

IQdiscr <- cut(IQ, quantile(IQ), include.lowest=TRUE)
summary(IQdiscr)

[70.5,92.6]  (92.6,102]   (102,108]   (108,127] 
          5           5           5           5 

Recode factors

Change names of factor levels

sex <- c(1, 0, 0, 1, 1, 1, 0, 0)
(sexFac <- factor(sex, levels=0:1, labels=c("f", "m")))

[1] m f f m m m f f
Levels: f m

levels(sexFac) <- c("female", "male")
sexFac

[1] male   female female male   male   male   female female
Levels: female male

Using package forcats

library(forcats)
fct_recode(sexFac, "F"="female", "M"="male")

[1] M F F M M M F F
Levels: F M

Add, combine and remove factor levels

Add factor levels

(status <- factor(c("hi", "lo", "hi")))

[1] hi lo hi
Levels: hi lo

status[4] <- "mid"
status

[1] hi   lo   hi   <NA>
Levels: hi lo

levels(status) <- c(levels(status), "mid")
status[4] <- "mid"
status

[1] hi  lo  hi  mid
Levels: hi lo mid

Using package forcats

fct_expand(status, "new_level")

[1] hi  lo  hi  mid
Levels: hi lo mid new_level

Combine factor levels

hiNotHi <- status
levels(hiNotHi) <- list(hi="hi", notHi=c("mid", "lo"))
hiNotHi

[1] hi    notHi hi    notHi
Levels: hi notHi

Using package forcats

fct_collapse(status, notHi=c("mid", "lo"))

[1] hi    notHi hi    notHi
Levels: hi notHi

Remove factor levels

status[1:2]

[1] hi lo
Levels: hi lo mid

(newStatus <- droplevels(status[1:2]))

[1] hi lo
Levels: hi lo

Using package forcats

fct_drop(status[1:2], "mid")

[1] hi lo
Levels: hi lo

Reorder factor levels

Using fct_relevel() from package forcats

(facGrp <- factor(rep(LETTERS[1:3], each=5)))

 [1] A A A A A B B B B B C C C C C
Levels: A B C

library(forcats)
fct_relevel(facGrp, "A", after=1)

 [1] A A A A A B B B B B C C C C C
Levels: B A C

fct_relevel(facGrp, "A", after=Inf)

 [1] A A A A A B B B B B C C C C C
Levels: B C A

Reorder group levels according to group statistics

vec <- rnorm(15, rep(c(10, 5, 15), each=5), 3)
tapply(vec, facGrp, FUN=mean)

        A         B         C 
 7.800560  4.652087 16.635740 

reorder(facGrp, vec, FUN=mean)

 [1] A A A A A B B B B B C C C C C
attr(,"scores")
        A         B         C 
 7.800560  4.652087 16.635740 
Levels: B A C

Detach (automatically) loaded packages (if possible)

try(detach(package:dplyr))
try(detach(package:forcats))

Get the article source from GitHub

R markdown - markdown - R code - all posts