CODEWITHSUNDEEP
rdata.table
Jeffrey Kramer
Jeffrey Kramer

Reputation: 1345

R rollapply throws errors in data.table

I have a data table and want to execute a rollapply to calculate a rolling average for a particular group. I have:

tt <- data.table(tt, key = c("dt", "id"))

dt      id  val
1/1/14  a    NA
1/2/14  a    3
1/3/14  a    2
1/4/14  a    5
...      
2/1/14  a    NA
1/1/14  b    NA
1/2/14  b    9
1/3/14  b   10
...
2/1/14  b   12

I want to do a three-day rolling average for each id whenever there is valid data. I tried:

tt[, test := rollapply(val, FUN = mean, 
  width = 3, na.rm = TRUE), by = list(id))

This gives some odd warnings and then NaNs. Have I done something wrong?

Upvotes: 1

Views: 432

Answers (1)

Jeffrey Kramer
Jeffrey Kramer

Reputation: 1345

This was resolved within the rollapply function itself. I needed to add the partial argument. This worked great:

tt[, test := rollapply(val, FUN = mean, width = 3, na.rm = TRUE, 
   partial = TRUE), by = list(id)]

Upvotes: 1

Related Questions