How to properly implement user-defined aggregate function with dcast

Question

I have a data frame like

df<-data.frame(date=c(rep("1/27/2010",times=30)),
           loc1=c(rep(9:13,each=6)),
           loc2=c(rep(c("N","E","W"),each=2)),
           loc3=c(rep(c(1,2))),
           tr1=c(rep(c(0,1),each=15)),
           tr2=c(0,0,0,0,0,0,0,1,1,1,1,1,1,1,1,0,0,0,0,0,0,0,0,1,1,1,1,1,1,1),
           tr3=c(1,1,1,1,1,1,1,2,2,2,2,2,2,2,2,3,3,3,3,3,3,3,4,4,4,4,4,4,4,4),
           Birth=c(sample(c("early","late"),30,replace=TRUE,prob=c(0.5,0.5))),
           Species=c(rep(c("A","B"),times=15)),
           Status=c(sample(c(0,1),30,replace=TRUE,prob=c(0.7,0.3))))

df<-rbind(df,df)

I want to make separate columns for each value of loc3, with rows defined by loc1,loc2,tr1,tr2,tr3,Birth, and Species. I want to 'count' the statuses of all the observations that share these values and group the counts by loc3.

I planned to use dcast from the reshape2 package.

I wrote a function to perform the 'count' I want. I'm new to R and while I'm sure there is a function that does this, I couldn't find it immediately and it seemed a worthwhile exercise to try and write the script myself, given the simplicity of the task.

      d.count<-function(x){
  j=0
  for (i in 1:length(x))
    if (is.na(x{i])){
      j<-j+0
    }else if(x[i]==0){
      j<-j+1
    } else if(x[i]==1){
      j<-j+0
    }
  return(j)
}

0s should increase the count and 1s and NAs shouldn't.

So

df_1<-dcast(df,date+loc1+loc2+tr1+tr2+tr3+Birth+Species~loc3,value.var="Status",fun.aggregate=d.count)

I get the error

Error in if (is.na(x[i])) { : argument is of length zero

Which makes me think I do not understand how dcast is treating fun.aggregate...

Thanks for the help! -JJE

dickoa dickoa · Accepted Answer · 2013-06-18T21:12:29

Why not something like this using the tabulate function

require(reshape2)
dcast(df, ... ~ loc3, value.var = "Status", fun.aggregate = tabulate)

##         date loc1 loc2 tr1 tr2 tr3 Birth Species 1 2
## 1  1/27/2010    9    E   0   0   1 early       A 0 0
## 2  1/27/2010    9    E   0   0   1 early       B 0 0
## 3  1/27/2010    9    N   0   0   1 early       B 0 0
## 4  1/27/2010    9    N   0   0   1  late       A 0 0
## 5  1/27/2010    9    W   0   0   1 early       B 0 0
## 6  1/27/2010    9    W   0   0   1  late       A 0 0
## 7  1/27/2010   10    E   0   1   2  late       A 0 0
## 8  1/27/2010   10    E   0   1   2  late       B 0 2
## 9  1/27/2010   10    N   0   0   1  late       A 0 0
## 10 1/27/2010   10    N   0   1   2  late       B 0 2
## 11 1/27/2010   10    W   0   1   2  late       A 0 0
## 12 1/27/2010   10    W   0   1   2  late       B 0 0
## 13 1/27/2010   11    E   0   1   2  late       A 0 0
## 14 1/27/2010   11    E   1   0   3 early       B 0 2
## 15 1/27/2010   11    N   0   1   2 early       B 0 0
## 16 1/27/2010   11    N   0   1   2  late       A 0 0
## 17 1/27/2010   11    W   1   0   3  late       A 0 0
## 18 1/27/2010   11    W   1   0   3  late       B 0 2
## 19 1/27/2010   12    E   1   0   3 early       B 0 0
## 20 1/27/2010   12    E   1   0   3  late       A 0 0
## 21 1/27/2010   12    N   1   0   3 early       A 2 0
## 22 1/27/2010   12    N   1   0   3 early       B 0 2
## 23 1/27/2010   12    W   1   0   4 early       A 0 0
## 24 1/27/2010   12    W   1   1   4 early       B 0 0
## 25 1/27/2010   13    E   1   1   4 early       B 0 0
## 26 1/27/2010   13    E   1   1   4  late       A 0 0
## 27 1/27/2010   13    N   1   1   4  late       A 0 0
## 28 1/27/2010   13    N   1   1   4  late       B 0 2
## 29 1/27/2010   13    W   1   1   4 early       A 0 0
## 30 1/27/2010   13    W   1   1   4 early       B 0 2

EDIT

If you want to count the number of 0 for example :

dcast(df, ... ~ loc3, value.var = "Status", 
         fun.aggregate = function(x) sum(x == 0, na.rm = TRUE))

How to properly implement user-defined aggregate function with dcast

1 Answers