Avoiding loops in R

Question

I have decided to learn R. I am trying to get a sense of how to write \"R style\" functions and to avoid looping. Here is a sample situation:

Given a vector a<

Answer 1

Because these rolling functions often apply with time-series data, some of the newer and richer time-series data-handling packages already do that for you:

R> library(zoo)   ## load zoo
R> speed <- c(1,1,1,1,1,4,6,3,6,8,9)
R> zsp <- zoo( speed, order.by=1:length(speed) )  ## creates a zoo object
R> rollmean(zsp, 5)                               ## default use
  3   4   5   6   7   8   9 
1.0 1.6 2.6 3.0 4.0 5.4 6.4 
R> rollmean(zsp, 5, na.pad=TRUE, align="right")   ## with padding and aligned
  1   2   3   4   5   6   7   8   9  10  11 
 NA  NA  NA  NA 1.0 1.6 2.6 3.0 4.0 5.4 6.4 
R> 

The zoo has excellent documentation that will show you many, many more examples, in particular how to do this with real (and possibly irregular) dates; xts extends this further but zoo is a better starting point.

Answer 2

You can also use a combination of cumsum and diff to get the sum over sliding windows. You'll need to pad with your own NaN, though:

> speed <- c(1,1,1,1,1,4,6,3,6,8,9)
> diff(cumsum(c(0,speed)), 5)/5
[1] 1.0 1.6 2.6 3.0 4.0 5.4 6.4

Answer 3

Something like b = filter(a, rep(1.0/5, 5), sides=1) will do the job, although you will probably get zeros in the first few slots, instead of NaN. R has a large library of built-in functions, and "R style" is to use those wherever possible. Take a look at the documentation for the filter function.