Search code examples
rdata.tablequantmod

R Finding the running correlation by group in data.table when enough rows does not exist

I am trying to find running correlation by group using runCor. Since the last Team has just one element, the runCor function is throwing the error.

library(data.table)
library(TTR)
dat <- data.table(
  Team = sapply(1:32, function(x) paste0("T", x)),
  Year = c(rep(c(2000:2009), 32)),
  Points_Game = c(rnorm(320, 100, 10))
)

dat = dat[order(Team, Year)][1:311]
# find correlation of Year and Points_Game for each Team
dat = dat[ , r := TTR::runCor(Year, Points_Game, n = 3), by = Team]

Is there a way to catch this case (T9 team) as correlation cannot be found and to fill NA in column r for T9?

Thanks!

Solution

  • You could use tryCatch to return NA in case of error:

    runCorProtected <- function(...) {
  tryCatch(TTR::runCor(...),error = function(e) {NA})
}

dat = dat[ , r := runCorProtected(Year, Points_Game, n = 3), by = Team]

dat[Team=='T9']
   Team Year Points_Game  r
1:   T9 2000    103.1672 NA