[R] How to do aggregate operations with non-scalar functions

[Itay Furman]

On Wed, 6 Apr 2005 james.holtman at convergys.com wrote:

> Here is a method that I use in this situation.  I work with the indices of
> the rows so that copies are not made and it is fast.
>
> Result <- lapply(split(seq(nrow(df)), df$a), function(.a){  # partition on
> the first variable
>      lapply(split(.a, df$z[.a]), function(.z){   # partition on the second
> variable -- notice the subsetting
>            c(quantile(df$r[.z]), ...anything else you want to compute)
>      })
> })
> Result <- do.call('rbind', Result)  # create a matrix - now you have your
> results
>
> Jim

Jim,

Thank you for your reply.  For some reason, when I try your 
proposed solution I get:

Error in sort(unique.default(x), na.last = TRUE) :
 	`x' must be atomic

Eventually, I used the solution proposed by Gabor G in this 
thread.  One advantage of his solution is that it is easier to 
scale up I believe;  for example in the case you have 3 factors 
that together subset the data frame.


 	Regards,
 	Itay

----------------------------------------------------------------
itayf at u.washington.edu  /  +1 (206) 543 9040  /  U of Washington