R help - Nov 2001 - inclusion criteria help

I have a dataset that looks like this (many other variables not
shown. including a unique row identifier "id"):
> summary(hits)
    query               lib               coverage         percid
 Length:80664       Length:80664       Min.   :0.080   Min.   :0.2250
 Mode  :character   Mode  :character   1st Qu.:0.980   1st Qu.:0.8160
                                       Median :1.000   Median :0.9230
                                       Mean   :0.946   Mean   :0.8536
                                       3rd Qu.:1.000   3rd Qu.:0.9900
                                       Max.   :1.000   Max.   :1.0000

For any query/lib combination there may be 1 or more rows of data. I'd
like to be able to specify only the rows for each query/lib combination
that have the maximum (or minimum or whatever) coverage or percid or some
other data element, and carry along the other corresponding data elements
from that same row.

I know I can do this procedurally in a loop:

query <- c('')
lib <- c('')
coverage <- c(0)
percid <- c(0)

for(q in unique(hits$query)) {
  for(l in unique(hits$lib[hits$query == q])) {
    query <- c(query, q)
    lib <- c(lib, l)
    max.coverage <- 0
    for(id in hits$id[hits$query == q & hits$lib == l]) {
      if(hits$coverage[hits$id == id] > max.coverage) {
        max.coverage.id <- id
        max.coverage <- hits$coverage[hits$id == id]
      }
    }
    coverage <- c(coverage, hits$coverage[hits$id == max.coverage.id])
    percid <- c(percid, hits$percid[hits$id == max.coverage.id])
  }
}

filtered.hits <- data.frame(query=query[2:length(query)],
                            lib=lib[2:length(lib)],
                            coverage=coverage[2:length(coverage)],
                            percid=percid[2:length(percid)]
                           )

# finally get to do something with it now:
plot(filtered.hits$coverage[filtered.hits$query == 'ABC'],
     filtered.hits$percid[filtered.hits$query == 'ABC']
    )


So, how could I accomplish the same plot as above without the looping and
creating a new dataframe?

Thanks,

-Aaron

-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at
stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._

On Tue, 27 Nov 2001, Aaron J Mackey wrote:
>
> I have a dataset that looks like this (many other variables not
> shown. including a unique row identifier "id"):
>
> > summary(hits)
>     query               lib               coverage         percid
>  Length:80664       Length:80664       Min.   :0.080   Min.   :0.2250
>  Mode  :character   Mode  :character   1st Qu.:0.980   1st Qu.:0.8160
>                                        Median :1.000   Median :0.9230
>                                        Mean   :0.946   Mean   :0.8536
>                                        3rd Qu.:1.000   3rd Qu.:0.9900
>                                        Max.   :1.000   Max.   :1.0000
>
> For any query/lib combination there may be 1 or more rows of data. I'd
> like to be able to specify only the rows for each query/lib combination
> that have the maximum (or minimum or whatever) coverage or percid or some
> other data element, and carry along the other corresponding data elements
> from that same row.
>
> I know I can do this procedurally in a loop:
>
<snip: he does it>
>
> So, how could I accomplish the same plot as above without the looping and
> creating a new dataframe?
Well, for one query you can do the subsetting like

	hits[hits$coverage == max(hits$coverage),]

so for many queries you could tapply() or by() this process

  filter<-function(this.subset){
	this.subset[this.subset$coverage==max(this.subset$coverage),]
	}
  filtered<-by(hits, hits$query,filter)

This produces a list of dataframes, so you want to staple them back
together
  filtered<-do.call("rbind",filtered)

In the case of maximum or minimum it would be faster to use the
which.max/which.min functions instead of the expression
    hits$coverage == max(hits$coverage)

This may or may not be faster than the loop, but it should be easier to
read (at least if you understand by()).


	-thomas


-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at
stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._