Tabulate a data frame in R

Question

I wanted to tabulate data so that a factor variable becomes columns and keep value from another variable in cell.

So I tried,

a=rep(1:3,3)
d<-rep(1:3, each=3)
b=rnorm(9)
c=runif(9)
dt<-data.frame(a,d,b,c)

  a d          b         c
1 1 1  0.3819762 0.5199602
2 2 1  0.3896063 0.9144730
3 3 1  2.4356972 0.2888464
4 1 2  1.2697016 0.9831191
5 2 2 -1.9844689 0.2046947
6 3 2  0.3473766 0.4766178
7 1 3 -1.5461235 0.6187189
8 2 3  1.0829027 0.9089551
9 3 3 -0.1305324 0.6326141

I looked for data.table, plyr, reshape2 but could not find what I wanted to do. So, I did the old loop way.

mat<-matrix(NA, nrow=3, ncol=4)


for (i in 1:3){
  mat[i,1]<-i
  for (j in 1:3){
    val=dt[a==i & d==j,3]
    mat[i,j+1]<-val

  }

}



mat
     [,1]      [,2]       [,3]       [,4]
[1,]    1 0.3819762  1.2697016 -1.5461235
[2,]    2 0.3896063 -1.9844689  1.0829027
[3,]    3 2.4356972  0.3473766 -0.1305324

... and it takes forever for big data.

Any better option??

Answer 1

Here's a data.table option:

library(data.table)
dt = data.table(dt)

dt[, as.list(b), by = a]

Answer 2

This can be done in base R also:

reshape(dt,timevar="d",idvar="a",drop="c",direction="wide")

For your data, this gives...

  a       b.1        b.2        b.3
1 1 0.3819762  1.2697016 -1.5461235
2 2 0.3896063 -1.9844689  1.0829027
3 3 2.4356972  0.3473766 -0.1305324

Please use set.seed before drawing simulated data, so that it is easier to reproduce.

I don't know that this solution will be fast. Also, to use it in the future, you have to get used to these confusing argument names ("timevar", "idvar", etc.) which probably don't describe what you're actually doing most of the time...

Answer 3

using reshape2

> library(reshape2)
> dcast(dt, a ~ d, value.var = "b")
  a         1          2          3
1 1 0.3819762  1.2697016 -1.5461235
2 2 0.3896063 -1.9844689  1.0829027
3 3 2.4356972  0.3473766 -0.1305324