Question

I’ve got a semi-long data frame in the following format:

| Id | Alt | Choice | Var1 | Var2 |
|----+-----+--------+------+------|
|  1 | a   | TRUE   | 58.2 |    4 |
|  1 | b   | FALSE  | 71.6 |    5 |
|  2 | a   | FALSE  | 14.8 |    2 |
|  2 | b   | TRUE   | 82.1 |    5 |

…that I’d like to reshape into the following format:

| Id | Choice.a | Choice.b | Var1.a | Var1.b | Var2.a | Var2.b |
|----+----------+----------+--------+--------+--------+--------|
| 1  | TRUE     | FALSE    | 58.2   | 71.6   | 4      | 5      |
| 2  | FALSE    | TRUE     | 14.8   | 82.1   | 2      | 5      |

I keep thinking to myself that I ought to be able to do something like:

library(reshape2)
dcast(df, ... ~ Alt)

But that’s not quite right.

Answer 1

This is not using ‘plyr’ but rather the base function reshape:

 dat <- read.table(text=" Id | Alt | Choice | Var1 | Var2 
  1 |a| TRUE   | 58.2 |    4 
  1 |b| FALSE  | 71.6 |    5 
  2 |a| FALSE  | 14.8 |    2
  2 |b| TRUE   | 82.1 |    5
 ", sep="|", header=TRUE)
 reshape(dat, direction="wide", idvar="Id", timevar="Alt")
#-----------------
  Id Choice.a Var1.a Var2.a Choice.b Var1.b Var2.b
1  1  TRUE      58.2      4  FALSE     71.6      5
3  2  FALSE     14.8      2  TRUE      82.1      5

If you need the precise order you showed, it’s easiy enough to use “[“. Imagine we assigned the result of reshape() to “wdat”:

wdat[c(1,2,5,3,6,4,7)]
#===============
  Id Choice.a Choice.b Var1.a Var1.b Var2.a Var2.b
1  1  TRUE     FALSE     58.2   71.6      4      5
3  2  FALSE    TRUE      14.8   82.1      2      5