One idiom for testing group-level conditions is:

data[, if (mean(x) < 10) .SD, by=g]

This might be slower in the special case of taking a mean. See ?GForce.

There's a request for an idiom like SQL HAVING over here:
https://github.com/Rdatatable/data.table/issues/788

--Frank

On Wed, Aug 16, 2017 at 4:44 PM, Bernstein, Elliot J <
[email protected]> wrote:

> Is there a way to subset a data table by the result of a grouped
> aggregation without adding an interim column to the table? For example, if
> I want to select all rows for which the group mean value of x is less than
> 10, I can do the following:
>
>
>
> data <- data.table(x = 1:20, g = rep(c("a", "b"), each = 10))
>
> data[, mean.x := mean(x), by = .(g)]
>
> data[mean.x < 10,]
>
>
>
> But I’m not really interested in “mean.x”. Can I do the same thing without
> adding it to the table?
>
>
>
> Thanks.
>
>
>
> - Elliot
>
> _______________________________________________
> datatable-help mailing list
> [email protected]
> https://lists.r-forge.r-project.org/cgi-bin/mailman/
> listinfo/datatable-help
>
_______________________________________________
datatable-help mailing list
[email protected]
https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/datatable-help

Reply via email to