Rollapply for time series

Question

I am trying to calculate the rolling 20 period historical volatility. I take the daily returns:

ret<-ROC(data1)

And then I use rollapply to

Answer 1

I recommend runner function in runner package to apply any R function on rolling windows. Below runner output identical to zoo.

library(runner)

runner(
  x, 
  function(x) sd(x),
  k = 20, 
  na_pad = TRUE
)

---

runner can reproduce zoo output and also have options like handling unequally spaced data and other (for more go to documentation and vignettes for more).

library(runner)
library(zoo)

x <- c(0.000000000, 0.005277045, 0.023622047, 0.002564103,-0.002557545, -0.020512821,
       0.007853403,-0.012987013,  0.007894737,  0.015665796,  0.000000000, -0.002570694,
       0.002577320, -0.015424165, 0.002610966,  0.010416667,  0.002577320,  0.015424165, 
       0.000000000, -0.002531646, -0.002538071, 0.030534351,  0.014814815, -0.007299270,
       -0.009803922, -0.012376238,  0.002506266, -0.015000000,-0.002538071,  0.002544529)

identical(
  runner(x, sd, k = 20, na_pad = TRUE),
  rollapply(x, 20, sd, fill = NA, align = "right")
)

# centered alignment
identical(
  runner(x, sd, k = 20, lag = -10, na_pad = TRUE),
  rollapply(x, 20, sd, fill = NA, align = "center")
)

Answer 2

You need to use align='right' instead of using the default which is align='center', or instead of using rollapply, use the rollapplyr wrapper which has align='right' as the default.

From ?rollapply:

align specifyies whether the index of the result should be left- or right-aligned or centered (default) compared to the rolling window of observations. This argument is only used if width represents widths.

Although, for this, personally, I'd use runSD from the TTR package because it uses compiled code and will be faster.

Either of these should do what you expect, but the second one will be faster.

library(zoo)
rollapply(x, 20, sd, fill=NA, align='right')

library(TTR)
runSD(x, 20)