R Apply() function on specific dataframe columns

Question

I want to use the apply function on a dataframe, but only apply the function to the last 5 columns.

B<- by(wifi,(wifi$Room),FUN=function(y){apply(y, 2, A)})

This applies A to all the columns of y

B<- by(wifi,(wifi$Room),FUN=function(y){apply(y[4:9], 2, A)})

This applies A only to columns 4-9 of y, but the total return of B strips off the first 3 columns... I still want those, I just don't want A applied to them.

wifi[,1:3]+B

also does not do what I expected/wanted.

The 'by' call is complicating this question. If it's relevant you should rewrite the question to clarify (what is wifi$Room?). I've ignored by in my answer below. — leif, Aug 29 '13 at 06:32
You could `cbind(y[1:3], ...)` to the result you are getting. — IRTFM, Aug 29 '13 at 06:46

score 121 · Answer 1 · edited Sep 21 '20 at 14:26

121

lapply is probably a better choice than apply here, as apply first coerces your data.frame to an array which means all the columns must have the same type. Depending on your context, this could have unintended consequences.

The pattern is:

df[cols] <- lapply(df[cols], FUN)

The 'cols' vector can be variable names or indices. I prefer to use names whenever possible (it's robust to column reordering). So in your case this might be:

wifi[4:9] <- lapply(wifi[4:9], A)

An example of using column names:

wifi <- data.frame(A=1:4, B=runif(4), C=5:8)
wifi[c("B", "C")] <- lapply(wifi[c("B", "C")], function(x) -1 * x)

edited Sep 21 '20 at 14:26

madx

6,723
4
55
59

answered Aug 29 '13 at 06:36

leif

3,003
1
19
9

2

A small correction: wifi <- data.frame(A=1:4, B=runif(4), C=5:8) – jcfaria Dec 22 '13 at 14:34
Could you be more explicit about how you created the [cols] vector? – Mox Apr 04 '18 at 21:41
@Mox you can just do `cols <- c("var1", "var2")` – cparmstrong Apr 17 '18 at 16:57
1

as alternative using dplyr avoiding the redundancy to repeat the column specification, you could do `wifi[4:9] %<>% map_dbl(A)` – Agile Bean Jan 16 '19 at 04:58
@AgileBean: `map` is a nice alternative but I would advice using the `%<>%` operator. Please scroll to the end of https://r4ds.had.co.nz/pipes.html – Kay Mar 14 '21 at 12:40

score 70 · Accepted Answer · answered Aug 29 '13 at 05:56

70

Using an example data.frame and example function (just +1 to all values)

A <- function(x) x + 1
wifi <- data.frame(replicate(9,1:4))
wifi

#  X1 X2 X3 X4 X5 X6 X7 X8 X9
#1  1  1  1  1  1  1  1  1  1
#2  2  2  2  2  2  2  2  2  2
#3  3  3  3  3  3  3  3  3  3
#4  4  4  4  4  4  4  4  4  4

data.frame(wifi[1:3], apply(wifi[4:9],2, A) )
#or
cbind(wifi[1:3], apply(wifi[4:9],2, A) )

#  X1 X2 X3 X4 X5 X6 X7 X8 X9
#1  1  1  1  2  2  2  2  2  2
#2  2  2  2  3  3  3  3  3  3
#3  3  3  3  4  4  4  4  4  4
#4  4  4  4  5  5  5  5  5  5

Or even:

data.frame(wifi[1:3], lapply(wifi[4:9], A) )
#or
cbind(wifi[1:3], lapply(wifi[4:9], A) )

#  X1 X2 X3 X4 X5 X6 X7 X8 X9
#1  1  1  1  2  2  2  2  2  2
#2  2  2  2  3  3  3  3  3  3
#3  3  3  3  4  4  4  4  4  4
#4  4  4  4  5  5  5  5  5  5

answered Aug 29 '13 at 05:56

thelatemail

91,185
12
128
188

Is there a way to do this using `$` to index a certain column by name instead of using `[ : ]` to index by column number? I tried adding colnames: `colnames(wifi) = c("a", "b", "c", "d", "e", "f", "g", "h" ,"i")` but any attempt at using lapply(wifi$e, 2, X) wasn't happening. – santeko Apr 21 '15 at 23:27
9

@skotturi - you can do this like `wifi[c("a","b","c")]` to index multiple columns by name. – thelatemail Apr 21 '15 at 23:35
@thelatemail,In `apply(wifi[4:9],2, A)`,`wifi[4:9]` is `data.frame`.And `apply` can only used to array or matrix.Why your answer workable? – kittygirl May 22 '20 at 18:18
@kittygirl - that's because apply *can* be used on a data.frame. The data.frame will be coerced to a matrix as part of the function when apply is used. – thelatemail May 22 '20 at 18:25
@thelatemail,will lose rowname or colname information? – kittygirl May 22 '20 at 18:28

score 5 · Answer 3 · answered Jan 25 '22 at 14:26

This task is easily achieved with the dplyr package's across functionality.

Borrowing the data structure suggested by thelatemail:

A <- function(x) x + 1
wifi <- data.frame(replicate(9,1:4))

We can indicate the columns we wish to apply the function to either by index like this:

library(dplyr)
wifi %>% 
   mutate(across(4:9, A))
#  X1 X2 X3 X4 X5 X6 X7 X8 X9
#1  1  1  1  2  2  2  2  2  2
#2  2  2  2  3  3  3  3  3  3
#3  3  3  3  4  4  4  4  4  4
#4  4  4  4  5  5  5  5  5  5

Or by name:

wifi %>% 
   mutate(across(X4:X9, A))
#  X1 X2 X3 X4 X5 X6 X7 X8 X9
#1  1  1  1  2  2  2  2  2  2
#2  2  2  2  3  3  3  3  3  3
#3  3  3  3  4  4  4  4  4  4
#4  4  4  4  5  5  5  5  5  5

score 1 · Answer 4 · answered Apr 28 '18 at 20:06

As mentioned, you simply want the standard R apply function applied to columns (MARGIN=2):

wifi[,4:9] <- apply(wifi[,4:9], MARGIN=2, FUN=A)

Or, for short:

wifi[,4:9] <- apply(wifi[,4:9], 2, A)

This updates columns 4:9 in-place using the A() function. Now, let's assume that na.rm is an argument to A(), which it probably should be. We can pass na.rm=T to remove NA values from the computation like so:

wifi[,4:9] <- apply(wifi[,4:9], MARGIN=2, FUN=A, na.rm=T)

The same is true for any other arguments you want to pass to your custom function.

score 0 · Answer 5 · edited Nov 10 '21 at 13:26

0

The easiest way is to use the mutate function:

dataFunctionUsed <- data %>% 
  mutate(columnToUseFunctionOn = function(oldColumn ...))

edited Nov 10 '21 at 13:26

Suraj Rao

29,388
11
94
103

answered Nov 10 '21 at 13:25

Henrik Mader

33
4

Mox · Answer 6 · 2018-04-04T21:46:38.227

-2

I think what you want is mapply. You could apply the function to all columns, and then just drop the columns you don't want. However, if you are applying different functions to different columns, it seems likely what you want is mutate, from the dplyr package.

edited Apr 04 '18 at 21:46

answered Apr 04 '18 at 21:40

Mox

511
5
15

R Apply() function on specific dataframe columns

6 Answers6

Linked

Related