applying rolling mean by group in R

Question

I\'m an R newbie and I\'m having a lot of trouble doing something that is probably very simple. I have a big dataset split up into groups by country code, and I want to take

Answer 1

In your first attempt, your function does not use its x argument, and always returns the same thing (a vector with the wrong size). In addition, the first argument, should be a vector. Lastly, tapply returns a list of vectors: you cannot put the result directly into a data.frame.

library(zoo)
n <- 10
leader <- data.frame(
  ccode = rep(LETTERS[1:3],each=n),
  GI_delta = rnorm(3*n)
)
tapply(
  leader$GI_delta, 
  leader$ccode, 
  function(x) rollmean(x, 3, na.pad=TRUE)
)

In your second example, the third argument of plyr should be a function, not an expression. If you want to use an expression, you can use summarize or transform as a function (summarize returns a 1-row data.frame for each value of ccode, while transform keeps the number of rows unchanged), and put the expressions as further arguments.

library(plyr)
ddply(
  leader, "ccode",
  transform,
  last3 = rollmean( GI_delta, 3, align="right", na.pad=TRUE )
)

Answer 2

If you want to make a new column, then try using ave. It resembles tapply but returns a vector of the same length as its first argument. My experience is that it is a lot faster than ddply:

require(zoo)
leader$last3<-ave(leader$GI_delta, leader$ccode, 
                         FUN= function(x) rollmean(x, k=3, na.pad=T) )