Calculate realized monthly values by quarter

Question

I'm stuck with a little R-problem. Let's assume I have a zoo data set that is structured as follows:

df<-data.frame(r1=rnorm(25), r2=rnorm(25))
df<-zoo(df, as.Date(seq(from=as.Date("1980-01-01"), to=as.Date("1983-02-01"), by="1 month")))

I would like to calculate the quarterly average value of both r1 and r2 per month, in the following way (in pseudo-code):

rx (Jan. 1980) = rx (Jan. 1980)
rx (Feb. 1980) = average [ rx (Jan. 1980), rx (Feb. 1980) ]
rx (Mar. 1980) = average [ rx (Jan. 1980), rx (Feb. 1980), rx (Mar. 1980)]

rx (Apr. 1980) = rx (Apr. 1980)
rx (May  1980) = average [ rx (Apr. 1980), rx (May  1980) ]
rx (Jun. 1980) = average [ rx (Apr. 1980), rx (May  1980), rx (Jun. 1980)]

etc. - that is, I would like to replace the value of each month by the value of the realized observations in the quarter up to this point in the quarter.

I've experimented with rollapply (for month 2 with parameters width=2, align = "right"; for month 3 with width=3), but I feel that either I can't figure out the smartest way to do it, or there's a better/faster way to do this... any suggestions would be highly appreciated!

Thanks, Philipp

Answer 1

Don't know if it's the best way, either. But it works! I experimented with rollapply, and finally got it working using a variable width parameter.

rollapply(df, ((month(index(df)) - 1) %% 3) + 1, mean, align="right")

Answer 2

Unless I'm missing something, a way is this:

fac <- as.numeric(cut(as.Date(attributes(df)$index, "%Y-%m-%d"), "quarter"))
split_df <- split(df, fac)                                                       
newdf <- do.call(rbind, lapply(split_df, 
                          function(x) {
                               x$r1 <- cumsum(x$r1) / seq_along(x$r1); 
                               x$r2 <- cumsum(x$r2) / seq_along(x$r2); 
                               return(x)}))
#newdf                                  #df
#                     r1           r2   #                    r1          r2
#1980-01-01 -0.056649139 -0.816007382   #1980-01-01 -0.05664914 -0.81600738
#1980-02-01  0.008543219 -0.423027620   #1980-02-01  0.07373558 -0.03004786
#1980-03-01  0.395468481 -0.755660995   #1980-03-01  1.16931901 -1.42092775
#1980-04-01 -0.375906206 -1.011203256   #1980-04-01 -0.37590621 -1.01120326
#1980-05-01 -0.131085288 -0.876251192   #1980-05-01  0.11373563 -0.74129913
#1980-06-01  0.025572095 -0.347781855   #1980-06-01  0.33888686  0.70915682