Divide observation by period mean. Help to simplify code - r

Question

Link to data:
http://dl.dropbox.com/u/56075871/data.txt
I want to divide each observation by mean for that hour. Example:
2012-01-02 10:00:00 5.23
2012-01-03 10:00:00 5.28
2012-01-04 10:00:00 5.29
2012-01-05 10:00:00 5.29
2012-01-09 10:00:00 5.28
2012-01-10 10:00:00 5.33
2012-01-11 10:00:00 5.42
2012-01-12 10:00:00 5.55
2012-01-13 10:00:00 5.68
2012-01-16 10:00:00 5.53
mean for that is 5.388. Next i want divide each observation by that mean, so... 5.23/5.388, 5.28/5.388, ... until end 5.53/5.388
I have hourly timeseries for 10 stocks:
S1.1h S2.1h S3.1h S4.1h S5.1h S6.1h S7.1h S8.1h S9.1h S10.1h
2012-01-02 10:00:00 64.00 110.7 5.23 142.0 20.75 34.12 32.53 311.9 7.82 5.31
2012-01-02 11:00:00 64.00 110.8 5.30 143.2 20.90 34.27 32.81 312.0 7.97 5.34
2012-01-02 12:00:00 64.00 111.1 5.30 142.8 20.90 34.28 32.70 312.4 7.98 5.33
2012-01-02 13:00:00 61.45 114.7 5.30 143.1 21.01 34.35 32.85 313.0 7.96 5.35
2012-01-02 14:00:00 61.45 116.2 5.26 143.7 21.10 34.60 32.99 312.9 7.95 5.36
2012-01-02 15:00:00 63.95 116.2 5.26 143.2 21.26 34.72 33.00 312.6 7.99 5.37
2012-01-02 16:00:00 63.95 117.3 5.25 143.3 21.27 35.08 33.04 312.7 7.99 5.36
2012-01-02 17:00:00 63.95 117.8 5.24 144.7 21.25 35.40 33.10 313.6 7.99 5.40
2012-01-02 18:00:00 63.95 117.9 5.23 145.0 21.20 35.50 33.17 312.5 7.98 5.35
2012-01-03 10:00:00 63.95 115.5 5.28 143.5 21.15 35.31 33.05 311.7 7.94 5.37
...
And i want to divie each observation by its mean for hour (periodical)
I have some code. Code to make means:
#10:00:00, 11:00:00, ... 18:00:00
times <- paste(seq(10, 18),":00:00", sep="")
#means - matrix of means for timeseries and hour
means <- matrix(ncol= ncol(time_series), nrow = length(times))
for (t in 1:length(times)) {
#t is time 10 to 18
for(i in 1:ncol(time_series)) {
#i is stock 1 to 10
# hour mean for each observation in data
means[t,i] <- mean(time_series[grep(times[t], index(time_series)), i])
}
}
And my function to get "things done":
for (t in 1:length(times)) {
# get all dates with times[t] hour
hours <- time_series[grep(times[t], index(time_series))]
ep <- endpoints(hours, "hours")
out <- rbind(out, period.apply(hours, INDEX=ep, FUN=function(x) {
x/means[t,]
}))
}
I know this is awful, but it works. How can i simplify code?

Answer 1

Here's one way to do it:
# Split the xts object into chunks by hour
# .indexhour() returns the hourly portion for each timestamp
s <- split(time_series, .indexhour(time_series))
# Use sweep to divide each value of x by colMeans(x) for each group of hours
l <- lapply(s, function(x) sweep(x, 2, colMeans(x), FUN="/"))
# rbind everything back together
r <- do.call(rbind, l)

Answer 2

The scale function can do that. Used with ave you could restrict to calcs within hours. Post the resutls of dput on that xts/zoo object and you will get rapid replies.