How do I apply a function in R to certain columns of a data frame grouped by another column?

Question

I've been looking at the help page for tapply and by and I'm not sure if they are the right tool for this. For example, if I have a dataframe where the columns are Name,Value1,Value2 and I want to apply a function, say function f(x,y) { do_something } to Value1 and Value2 grouped by Name and get as a result a dataframe with the columns Name,f(Value1,Value2) how should I go about that?

I can get tapply to work in a simple case like this:

tapply(df$Name, df$value1, mean)

but what if my function takes as input df$value2 as well? and is not as simple as mean? In other words, pseudo-notation for what I'm trying to do would be:

tapply(df$Name, c(df$value1,df$value2), function f(x,y) { x+y+bla...})

Answer 1

by will do the job although it will not return a data.frame.

by(df, df$Name, function(X) f(X$Value1, X$Value2))

The package data.table is better set up for this sort of thing:

install.packages("data.table")
library(data.table)
dt = data.table(df)
dt[,f(Value1, Value2),by=Name]

Will return exactly what you're looking for.

Answer 2

Also check out plyr. For example

require(plyr)
ddply(mtcars, .variables="cyl", .fun=mutate,
      meaningless_number = mean(mpg) + disp)

will give you back a data frame just like mtcars, with the added column meaningless_number which is the mean mpg by cyl plus the individual disp. Use .fun = mutate to add columns, .fun = summarize to see summaries, and other functions for other purposes.

The answers to this question are very good for general *apply knowledge. I also found this answer to be a great plyr tutorial.