How to use sapply to divide different rows for each column

Question

I'm a bit new at R, but I am trying to do something very simple using sapply because I will need to do it a lot. Say that you have many variables for 5 years, and want to divide the fifth row's values by the first row's for each of the columns at once.

a   b   c     
184 20  55
100 32  563
18  12  88
5   99  52
32  36  22

So far I can either do it one by one:

df$a<-(df[5,]$a/df[1,]$a)

Or if I try to use sapply:

df2<-data.frame(sapply(names(df)[-1], function(x) {
   (df[x]/df[x])
   }))

The problem is that I don't know how to denote the rows with the sapply so above I'm just dividing vars by themselves. What's the quickest way of doing this? Thanks!

markus · Answer 1 · 2018-08-20T21:05:30.650

0

For this task you don't need sapply and instead do

df[5, ] <- df[5, ] / df[1, ]
df
#           a    b     c
#1 184.000000 20.0  55.0
#2 100.000000 32.0 563.0
#3  18.000000 12.0  88.0
#4   5.000000 99.0  52.0
#5   0.173913  1.8   0.4

Referring to @Mako212's comment, if your data contains non-numeric columns you could create a logical vector first that is TRUE at positions where you're data contains numeric columns. Use it for column subsetting and then do the operation.

col_idx <- sapply(df, is.numeric)
df[5, col_idx] <- df[5, col_idx] / df[1, col_idx]

edited Aug 20 '18 at 21:05

answered Aug 07 '18 at 21:29

markus

25,843
5
39
58

1

This will throw an error if you have non-numeric columns, a simple way to avoid that would be to select your numeric columns `iris[1,1:4]/iris[5,1:4]`. – Mako212 Aug 07 '18 at 21:31
Yes, but OP does not mention non-numeric columns and writes "for each of the columns at once". – markus Aug 07 '18 at 21:32
Thank you! This seemed easiest for me, and i ended up selecting which columns are numeric because I was getting an error with this bit col_idx <- sapply(iris, is.numeric). It would be nice to not have to specify tho. – Mia Aug 20 '18 at 16:20
@Mia Thanks for the reply. if the name of your data set is `df`, then do `col_idx <- sapply(df, is.numeric)`. I used the `iris` data for illustration only since you didn't mention non-numeric columns in your question. – markus Aug 20 '18 at 21:05

score 0 · Answer 2 · answered Aug 07 '18 at 21:56

If you have mixed type columns here is a dplyr approach

library(dplyr)
df %>% mutate_if(is.numeric, function(x) replace(x, length(x), x[length(x)] / x[1]))
#           a    b     c d e
#1 184.000000 20.0  55.0 a A
#2 100.000000 32.0 563.0 b B
#3  18.000000 12.0  88.0 c C
#4   5.000000 99.0  52.0 d D
#5   0.173913  1.8   0.4 e E

Sample data

# Sample data with mixed numeric and character columns
df <- read.table(text =
    "a   b   c
184 20  55
100 32  563
18  12  88
5   99  52
32  36  22 ", header = T)
df <- cbind(df, d = letters[1:5], e = LETTERS[1:5])

score 0 · Answer 3 · answered Aug 07 '18 at 22:12

This is something that could be good as a function:

library(dplyr)

div_row <- function(data, numerator, denominator){
    data %>% mutate_if(is.numeric, funs(if_else(row_number() == numerator, .[numerator]/.[denominator], .)))
}

df %>% div_row(5,1)
#     a      b     c    d    
# 1 184      20    55   a
# 2 100      32   563   a    
# 3  18      12    88   c    
# 4   5      99    52   e    
# 5   0.174   1.8   0.4 t  

df %>% div_row(2,1)
#     a      b     c    d 
# 1 184      20    55   a    
# 2   0.543   1.6  10.2 a 
# 3  18      12    88   c
# 4   5      99    52   e  
# 5  32      36    22   t

How to use sapply to divide different rows for each column

3 Answers3

Sample data