Averaging over continuous blocks

I have a data like this:

f  x 
A 1.1
A 2.2
A 3.3
B 3.5
B 3.7
B 3.9
B 4.1
B 4.5
A 5.1
A 5.2
C 5.4
C 5.5
C 6.1
B 6.2
B 6.3

I would like to average x over continuous blocks of f, to get this, similar to tapply(...,mean), but aware of the fact that it shouldn't mix separated blocks and in origi开发者_StackOverflownal order:

f  x
A 2.2
B 3.94
A 5.15 
C 5.67
B 6.25

---

Here's one way:

## reproducible code for example
dat <- read.table(foo <- textConnection("f  x 
A 1.1
A 2.2
A 3.3
B 3.5
B 3.7
B 3.9
B 4.1
B 4.5
A 5.1
A 5.2
C 5.4
C 5.5
C 6.1
B 6.2
B 6.3
"), header = TRUE)
close(foo)

We use rle() to compute the run lengths of f and the create a new factor fac that indexes the changes, for want of a better word, in f. We then aggregate on f and fac:

lens <- with(dat, rle(as.character(f)))
dat$fac <- with(lens, factor(rep(seq_along(lengths), times = lengths)))
aggregate(x ~ f + fac, data = dat, FUN = mean)

Giving:

> aggregate(x ~ f + fac, data = dat, FUN = mean)
  f fac        x
1 A   1 2.200000
2 B   2 3.940000
3 A   3 5.150000
4 C   4 5.666667
5 B   5 6.250000

We can easily drop the second column fac in the result if that is undesirable:

> aggregate(x ~ f + fac, data = dat, FUN = mean)[,-2]
  f        x
1 A 2.200000
2 B 3.940000
3 A 5.150000
4 C 5.666667
5 B 6.250000

---

rle is one possibility :

> id <- rle(as.character(Data$f))
> Means <-tapply(Data$x,rep(1:length(id$lengths),id$lengths),mean)    
> data.frame(Means,f=id$values)
     Means f
1 2.200000 A
2 3.940000 B
3 5.150000 A
4 5.666667 C
5 6.250000 B

It gives you the runs and the values, so you can use both.