8

I am using dplyr and I am wondering whether it is possible to compute differences between groups in one line. As in the small example below, the task is to compute the difference between groups A and Bs standardized "cent" variables. 

library(dplyr)
# creating a small data.frame
GROUP <- rep(c("A","B"),each=10)
NUMBE <- rnorm(20,50,10)
datf <- data.frame(GROUP,NUMBE)

datf2 <- datf %.% group_by(GROUP) %.% mutate(cent = (NUMBE - mean(NUMBE))/sd(NUMBE))

gA <- datf2 %.% ungroup() %.% filter(GROUP == "A") %.% select(cent)
gB <- datf2 %.% ungroup() %.% filter(GROUP == "B") %.% select(cent)

gA - gB

This is of course no problem by creating different objects - but is there a more "built in" way of performing this task? Something more like this not working fantasy code below?

datf2 %.% summarize(filter(GROUP == "A",select(cent)) - filter(GROUP == "B",select(cent)))

Thank you!

fish2000
  • 4,289
  • 2
  • 37
  • 76
Manuel
  • 265
  • 3
  • 11
  • I think part of the problem here is that you are trying to subtract two vectors and the DF has no code for which elements "line up" so to speak. I think if you had a column of `c(1:10,1:10)` then grouped on that and did a `summarize` with a substract somewhere it might be doable #vague – Spacedman Mar 23 '14 at 12:33

2 Answers2

6

Given we have 10 of each group, add an index 1:10, 1:10 and summarize over that with difference:

> datf2$entry=c(1:10,1:10)
> datf2 %.% ungroup() %.% group_by(entry) %.% summarize(d=cent[1]-cent[2])
Source: local data frame [10 x 2]

   entry          d
1      1 -0.8272879
2      2 -0.9159827
3      3 -0.5064762
4      4  0.4211639
5      5  1.3681720
6      6  3.3430289
7      7  1.0086822
8      8 -0.6163907
9      9 -0.7325220
10    10 -2.5423875

compare:

> gA - gB
         cent
1  -0.8272879
2  -0.9159827
3  -0.5064762
4   0.4211639
5   1.3681720
6   3.3430289
7   1.0086822
8  -0.6163907
9  -0.7325220
10 -2.5423875

Is there a way to inject the entry field into the data or the dplyr call? I'm not sure, it seems to rely on the functions knowing too much about the data...

Spacedman
  • 92,590
  • 12
  • 140
  • 224
4

Thank you for the inspiration. I further developed this solution to that:

mutate(datf2,diffence = filter(datf2, GROUP == "A")$cent - filter(datf2, GROUP == "B")$cent)

This adds the result as column in the the data.frame.

Manuel
  • 265
  • 3
  • 11