[Fwd: Re: [R] vectorization of a data-aggregation loop]

[Christoph Lehmann]

great! many thanks, Phil

Cheers
christoph

Phil Spector wrote:
> Christoph -
>    I think reshape is the function you're looking for:
> 
>> tt <- data.frame(cbind(c(1,1,1,1,1,2,2,2,3,3,3,3),
> + c(10,12,8,33,34,3,27,77,34,45,4,39), c('a', 'b', 'b', 'a', 'c', 'c', 'c',
> + 'a', 'b', 'a', 'b', 'c')))
>>  reshape(aggregate(as.numeric(tt$iwv),list(id=tt$id,type=tt$type),sum),idvar="id",timevar="type",direction="wide") 
>>
>   id x.a x.b x.c
> 1  1   6  13   6
> 2  2  10  NA   7
> 3  3   9  14   7
> 
>                                        - Phil Spector
>                      Statistical Computing Facility
>                      Department of Statistics
>                      UC Berkeley
>                      spector at stat.berkeley.edu
> 
> 
> On Tue, 1 Feb 2005, Christoph Lehmann wrote:
> 
>> Hi
>> I have a simple question:
>>
>> the following data.frame
>>
>>   id iwv type
>> 1   1   1    a
>> 2   1   2    b
>> 3   1  11    b
>> 4   1   5    a
>> 5   1   6    c
>> 6   2   4    c
>> 7   2   3    c
>> 8   2  10    a
>> 9   3   6    b
>> 10  3   9    a
>> 11  3   8    b
>> 12  3   7    c
>>
>> shall be aggregated into the form:
>>
>>  id t.a t.b t.c
>> 1  1   6  13   6
>> 6  2  10   0   7
>> 9  3   9  14   7
>>
>> means for each 'type' (a, b, c) a new column is introduced which
>> gets the sum of iwv for the respective observations 'id'
>>
>> of course I can do this transformation/aggregation in a loop (see 
>> below), but is there a way to do this more efficiently, eg. in using 
>> tapply (or something similar)- since I have lot many rows?
>>
>> thanks for a hint
>>
>> christoph
>>
>> #------------------------------------------------------------------------------ 
>>
>> # the loop-way
>> t <- data.frame(cbind(c(1,1,1,1,1,2,2,2,3,3,3,3), 
>> c(10,12,8,33,34,3,27,77,34,45,4,39), c('a', 'b', 'b', 'a', 'c', 'c', 
>> 'c', 'a', 'b', 'a', 'b', 'c')))
>> names(t) <- c("id", "iwv", "type")
>> t$iwv <- as.numeric(t$iwv)
>> t
>>
>> # define the additional columns (type.a, type.b, type.c)
>> tt <- rep(0, nrow(t) * length(levels(t$type)))
>> dim(tt) <- c(nrow(t), length(levels(t$type)))
>> tt <- data.frame(tt)
>> dimnames(tt)[[2]] <- paste("t.", levels(t$type), sep = "")
>> t <- cbind(t, tt)
>> t
>>
>> obs <- 0
>> obs.previous <- 0
>> row.elim <- rep(FALSE, nrow(t))
>> ta <- which((names(t) == "t.a")) #number of column which codes the 
>> first type
>> r.ctr <- 0
>> for (i in 1:nrow(t)){
>>  obs <- t[i,]$id
>>  if (obs == obs.previous) {
>>    row.elim[i] <- TRUE
>>    r.ctr <- r.ctr + 1 #increment
>>    type.col <- as.numeric(t[i,]$type)
>>    t[i - r.ctr, ta - 1 + type.col] <- t[i - r.ctr, ta - 1 +
>>      type.col] + t[i,]$iwv
>>  }
>>  else {
>>    r.ctr <- 0 #record counter
>>    type.col <- as.numeric(t[i,]$type)
>>    t[i, ta - 1 + type.col] <- t[i,]$iwv
>>  }
>>  obs.previous <- obs
>> }
>>
>> t <- t[!row.elim,]
>> t <- subset(t, select = -c(iwv, type))
>> t
>>
>> #------------------------------------------------------------------------------ 
>>
>>
>> ______________________________________________
>> R-help at stat.math.ethz.ch mailing list
>> https://stat.ethz.ch/mailman/listinfo/r-help
>> PLEASE do read the posting guide! 
>> http://www.R-project.org/posting-guide.html
>>
> 
>