cov/cor between two metrics

[JianyangZhao]

For example like
(-1 * correlation(open, volume, 10))
If we have two metrics for open and volume each, and each column for a symbol. It will be much more convenient if we can use something like "roll_cor(open, volume, 10)" or "roll_cor(list(open, volume), 10)" rather than transform into vectors then transform back.

[jjf234]

Thanks for the feedback. Just so I understand what you’re thinking, here’s an example of how I’d approach it using a for loop in R:

library(roll)
library(xts)

n_lists <- 2
n_vars <- 10
n_obs <- 1000
start_date <- as.Date("2013-12-31")
dates <- seq(start_date, length.out = n_obs, by = "day")
tickers <- paste("Ticker", 1:n_vars, sep = "")

# create dummy data
data <- list()
for (i in 1:n_lists) {
  temp <- xts(matrix(rnorm(n_obs * n_vars), nrow = n_obs, ncol = n_vars), dates)
  colnames(temp) <- tickers
  data <- c(data, list(temp))
}

# rolling correlation
result <- list()
for (i in 1:n_vars) {
  temp <- do.call(merge, c(lapply(1:n_lists, function(j) data[[j]][ , i])))
  result <- c(result, list(roll_cor(temp, 252)))
}
names(result) <- tickers

result[["Ticker1"]][ , , n_obs]
result[["Ticker2"]][ , , n_obs]

If so, in addition to price vs volume for many securities, do you have other use cases that come to mind? My sense is more basic applications, such as when there is only one x (e.g. correlation of many securities to one market) or one y (e.g. correlation of one security to many markets), are more common and the approach above would not work.

[uzhao]

Current i use a similar implement. But if we can have a native cpp solution could it be more efficient in time cost and memory cost. For now we have to use a 2*2 matrix for saving 1 value.

roll_cov2 = function(x, y, n) {
ret = matrix(roll_cov(cbind(as.vector(x), as.vector(y)), n)[1,2,], nrow = nrow(x))
ret[1:(n-1),] = NA
ret
}

[jjf234]

Thanks, I believe you're simply asking to have the same behavior as in stats::cov(x, y = NULL). Makes sense and it's something I'd also like to include in future versions.

[jjf234]

This feature was added in the development version:

# install.packages("devtools")
devtools::install_github("jjf234/roll")