Message-ID: <422C8F3A.1070501@stat.wisc.edu>
Date: 2005-03-07T17:28:26Z
From: Douglas Bates
Subject: Faster way of binding multiple rows of data than rbind?
In-Reply-To: <BAY101-F17BAD384690B07F80C4364E85F0@phx.gbl>
Ken Termiso wrote:
> Hi all,
>
> I have a vector that contains the row numbers of data taken from several
> filtering operations performed on a large data frame (20,000rows x
> 500cols).
>
> In order to output this subset of data, I've been looping through the
> vector containing the row numbers (keepRows).
>
> output <- data.frame(row.names = rownames(bigMatrix))
>
> for(i in keepRows)
> {
> output <- rbind(output, bigMatrix[i, ])
> }
>
>
> As you may guess, doing all of these rbinds takes a LOT of time, so I'm
> wondering if there's a workaround where I can maybe use an intermediate
> matrix-like object to store the loop output, and then coerce it back to
> a data frame after the loop is complete??
The indexing operations in R are very flexible. You can do this in a
single operation as
output <- bigMatrix[keepRows, ]