[R] Reshaping data

[Peter Dalgaard]

"Rau, Roland" <Rau at demogr.mpg.de> writes:

> Dear all,
> 
> given I have data in a data.frame which indicate the number of people in
> a 
> specific year at a specific age:
> 
> n <- 10
> mydf <- data.frame(yr=sample(1:10, size=n, replace=FALSE),
>                    age=sample(1:12, size=n, replace=FALSE),
>                    no=sample(1:10, size=n, replace=FALSE))
> 
> Now I would like to make a matrix with (in this simple example)
> 10 columns (for the years) and 12 rows (for the ages). In each cell,
> I would like to put the correct number of individuals.
> 
> So far I was doing this as follows:
> 
> mymatrix <- matrix(0, ncol=10, nrow=12)
> for (year in unique(mydf$yr)) {
>   for (age in unique(mydf$age)) {
>     if (length(mydf$no[mydf$yr==year & mydf$age==age]) > 0) {
>       mymatrix[age,year] <- mydf$no[mydf$yr==year & mydf$age==age]
>     } else {
>       mymatrix[age,year] <- 0
>     }
>   }
> }
> 
> This is fairly fast in such a simple setting.
> But with more years and ages (and for roughly 300 datasets) this becomes
> pretty slow. And in addition, this is not really elegant R-code.
> 
> Can somebody point me into the direction how I can do that in a more
> elegant
> way, possibly avoiding the loops?

This almost gets you there:

with(mydf, tapply(no,list(age,yr), sum))

except that it puts NA where you want 0, which you could fix with

 m <- with(mydf, tapply(no,list(age,yr), sum))
 m[is.na(m)] <- 0
 m

Other options include matrix indexing:

with(mydf, {
  M <- matrix(0,12,10)
  M[cbind(age,yr)]<-no
})

or (tada...) the reshape() function, esp. if you want a data frame as
output.
-- 
   O__  ---- Peter Dalgaard             Øster Farimagsgade 5, Entr.B
  c/ /'_ --- Dept. of Biostatistics     PO Box 2099, 1014 Cph. K
 (*) \(*) -- University of Copenhagen   Denmark          Ph:  (+45) 35327918
~~~~~~~~~~ - (p.dalgaard at biostat.ku.dk)                  FAX: (+45) 35327907