Best way to make factor levels uniform over a number of data frames

Question

I have a number of data.frames that each have a factor. I want to make sure that they all use the same levels. What is the proper way to do this?

In the code below you'll see that I reassign the factor for each case using levels from the overall set of levels with a small convenience function changeLevels. I would expect that there is a better way to do this though.

set.seed(1234)
b<-round(runif(100,1,10),digits=2)
set.seed(2345)
b2<-round(runif(100,11,20),digits=2)
set.seed(3456)
b3<-round(runif(50,15,18),digits=2)

#.. all potential levels
bt<-factor(sort(c(b,b2,b3)))
lvls<-levels(bt)

t1<-as.data.frame(table(sample(b,5)))
t2<-as.data.frame(table(sample(b,1)))
t3<-as.data.frame(table(sample(b,1)))
t4<-as.data.frame(table(sample(b,8)))
t5<-as.data.frame(table(sample(b2,20)))
t6<-as.data.frame(table(sample(b3,18)))

t1<-cbind(t1,p="A")
t2<-cbind(t2,p="B")
t3<-cbind(t3,p="C")
t4<-cbind(t4,p="D")
t5<-cbind(t5,p="E")
t6<-cbind(t6,p="F")

d<-data.frame()
d<-rbind(d,t2,t3,t6,t4,t5,t1)

#.. out of order bins
ggplot(d,aes(x=factor(Var1),fill=factor(p))) +
  geom_bar(aes(weight=Freq)) +
  facet_grid( p ~ ., margins=T)+
  ggtitle("out of order bins")

changeFactor<-function(t,lvls){
  temp<-as.numeric(as.character(t))
  factor(temp,levels=lvls)
}

t1$Var1<-changeFactor(t1$Var1,lvls)
t2$Var1<-changeFactor(t2$Var1,lvls)
t3$Var1<-changeFactor(t3$Var1,lvls)
t4$Var1<-changeFactor(t4$Var1,lvls)
t5$Var1<-changeFactor(t5$Var1,lvls)
t6$Var1<-changeFactor(t6$Var1,lvls)

d<-data.frame()
d<-rbind(d,t2,t3,t6,t4,t5,t1)

#.. in order bins
ggplot(d,aes(x=factor(Var1),fill=factor(p))) +
  geom_bar(aes(weight=Freq)) +
  facet_grid( p ~ ., margins=T)+
  ggtitle("in order bins")

Answer 1

short answer: keep your data in lists and learn the *pply family

set.seed(1234)
b<-round(runif(100,1,10),digits=2)
set.seed(2345)
b2<-round(runif(100,11,20),digits=2)
set.seed(3456)
b3<-round(runif(50,15,18),digits=2)

#.. all potential levels
bt<-factor(sort(c(b,b2,b3)))
lvls<-levels(bt)

options(stringsAsFactors = FALSE)
f <- function(x, y, z)
  cbind(data.frame(table(sample(x, y))), p = z)

datl <- Map(f, list(b,b,b,b,b2,b3), c(5,1,1,8,20,18), LETTERS[1:6])

changeFactor<-function(t,lvls){
  temp<-as.numeric(as.character(t))
  factor(temp,levels=lvls)
}

datl <- lapply(rapply(datl, f = function(x) changeFactor(x, lvls), 
                     classes = 'factor', how = 'replace'),
              data.frame)

d <- do.call(rbind, datl[c(2, 3, 6, 4, 5, 1)])

#.. in order bins
ggplot(d,aes(x=factor(Var1),fill=factor(p))) +
  geom_bar(aes(weight=Freq)) +
  facet_grid( p ~ ., margins=T)+
  ggtitle("in order bins")

long answer:

set.seed(1234)
b<-round(runif(100,1,10),digits=2)
set.seed(2345)
b2<-round(runif(100,11,20),digits=2)
set.seed(3456)
b3<-round(runif(50,15,18),digits=2)

#.. all potential levels
bt<-factor(sort(c(b,b2,b3)))
lvls<-levels(bt)

first, I don't want any unexpected factors popping up, so stringsAsFactors = FALSE then write a function, f, to do what you want, and check to make sure it works

options(stringsAsFactors = FALSE)
f <- function(x, y, z)
  cbind(data.frame(table(sample(x, y))), p = z)

f(b, 5, 'A')

#   Var1 Freq p
# 1 1.13    1 A
# 2 1.46    1 A
# 3 2.09    1 A
# 4  2.5    1 A
# 5 7.02    1 A

seems to work, so just Map it to lists of arguments and check the output

datl <- Map(f, list(b,b,b,b,b2,b3), c(5,1,1,8,20,18), LETTERS[1:6])

# List of 6
# $ :'data.frame':  5 obs. of  3 variables:
#   ..$ Var1: Factor w/ 5 levels "2.02","3.09",..: 1 2 3 4 5
#   ..$ Freq: int [1:5] 1 1 1 1 1
#   ..$ p   : chr [1:5] "A" "A" "A" "A" ...
# $ :'data.frame':  1 obs. of  3 variables:
#   ..$ Var1: Factor w/ 1 level "1.63": 1
#   ..$ Freq: int 1
#   ..$ p   : chr "B"

so combine everything to use with ggplot

d <- do.call(rbind, datl[c(2, 3, 6, 4, 5, 1)])

library(ggplot2)
#.. out of order bins
ggplot(d,aes(x=factor(Var1),fill=factor(p))) +
  geom_bar(aes(weight=Freq)) +
  facet_grid( p ~ ., margins=T)+
  ggtitle("out of order bins")

changeFactor<-function(t,lvls){
  temp<-as.numeric(as.character(t))
  factor(temp,levels=lvls)
}

again making sure the function does what it is supposed to do on one data frame

changeFactor(datl[[1]]$Var1, lvls)

# [1] 2.02 3.09 3.79 3.89 8.3 
# 234 Levels: 1.09 1.12 1.13 1.24 1.36 1.38 1.41 1.46 1.63 1.66 1.81 1.95 ... 19.86

so apply it again to them all at once and check the output

datl <- lapply(rapply(datl, f = function(x) changeFactor(x, lvls), 
                     classes = 'factor', how = 'replace'),
              data.frame)
str(datl)
# List of 6
# $ :'data.frame':  5 obs. of  3 variables:
#   ..$ Var1: Factor w/ 234 levels "1.09","1.12",..: 13 28 41 45 81
#   ..$ Freq: int [1:5] 1 1 1 1 1
#   ..$ p   : chr [1:5] "A" "A" "A" "A" ...
# $ :'data.frame':  1 obs. of  3 variables:
#   ..$ Var1: Factor w/ 234 levels "1.09","1.12",..: 9
#   ..$ Freq: int 1
#   ..$ p   : chr "B"
# ...

combine again and plot

d <- do.call(rbind, datl[c(2, 3, 6, 4, 5, 1)])

#.. in order bins
ggplot(d,aes(x=factor(Var1),fill=factor(p))) +
  geom_bar(aes(weight=Freq)) +
  facet_grid( p ~ ., margins=T)+
  ggtitle("in order bins")