16

I have a relational dataset, where I'm looking for dyadic information.

I have 4 columns. Sender, Receiver, Attribute, Edge

I'm looking to take the repeated Sender -- Receiver counts and convert them as additional edges.

df <- data.frame(sender = c(1,1,1,1,3,5), receiver = c(1,2,2,2,4,5), 
                attribute = c(12,12,12,12,13,13), edge = c(0,1,1,1,1,0))

   sender receiver attribute edge
1       1        1        12    0
2       1        2        12    1
3       1        2        12    1
4       1        2        12    1
5       3        4        13    1

I want the end result to look like this:

  sender receiver attribute edge
1      1        1        12    0
2      1        2        12    3
3      3        4        13    1

Where the relationship between duplicate sender-receivers have been combined and the number of duplicates incorporated in the number of edges.

Any input would be really appreciated.

Thanks!

Matt Dowle
  • 58,872
  • 22
  • 166
  • 224
crock1255
  • 1,025
  • 2
  • 12
  • 23

2 Answers2

20

For fun, here are two other options, first using the base function aggregate() and the second using data.table package:

> aggregate(edge ~ sender + receiver + attribute, FUN = "sum", data = df)
  sender receiver attribute edge
1      1        1        12    0
2      1        2        12    3
3      3        4        13    1
4      5        5        13    0
> require(data.table)
> dt <- data.table(df)
> dt[, list(sumedge = sum(edge)), by = "sender, receiver, attribute"]
     sender receiver attribute sumedge
[1,]      1        1        12       0
[2,]      1        2        12       3
[3,]      3        4        13       1
[4,]      5        5        13       0

For the record, this question has been asked many many many times, perusing my own answers yields several answers that would point you down the right path.

Community
  • 1
  • 1
Chase
  • 67,710
  • 18
  • 144
  • 161
7

plyr is your friend - although I think your end result is not quite correct given the input data. 

library(plyr)

ddply(df, .(sender, receiver, attribute), summarize, edge = sum(edge))

Returns

  sender receiver attribute edge
1      1        1        12    0
2      1        2        12    3
3      3        4        13    1
4      5        5        13    0
mnel
  • 113,303
  • 27
  • 265
  • 254
  • I think the OP was not intending to group by `sender + receiver + attribute`, but just by `sender + receiver`, and `attribute` goes along for the ride. In the example, `attribute` just happens to be unique for the `sender + receiver` pairings, but I *think* that was accidental – Mark Lakata Nov 30 '16 at 21:32