1

I wanted to create new columns for my data.table based on ratio calculation. The names of my variables are slightly in a standard way so I think there must be a way to easily achieve this in data.table. However I am not able to get how to achieve this. Below is my sample data and code -

set.seed(1200)

ID <- seq(1001,1100)
region <- sample(1:10,100,replace = T)
Q21 <- sample(1:5,100,replace = T)
Q22 <- sample(1:15,100,replace = T)
Q24_LOC_1 <- sample(1:8,100,replace = T)
Q24_LOC_2 <- sample(1:8,100,replace = T)
Q24_LOC_3 <- sample(1:8,100,replace = T)
Q24_LOC_4 <- sample(1:8,100,replace = T)

Q21_PAN <- sample(1:5,100,replace = T)
Q22_PAN <- sample(1:15,100,replace = T)
Q24_LOC_1_PAN <- sample(1:8,100,replace = T)
Q24_LOC_2_PAN <- sample(1:8,100,replace = T)
Q24_LOC_3_PAN <- sample(1:8,100,replace = T)
Q24_LOC_4_PAN <- sample(1:8,100,replace = T)

df1 <- as.data.table(data.frame(ID,region,Q21,Q22,Q24_LOC_1,Q24_LOC_2,Q24_LOC_3,Q24_LOC_4,Q21_PAN,Q22_PAN,Q24_LOC_1_PAN,Q24_LOC_2_PAN,Q24_LOC_3_PAN,Q24_LOC_4_PAN))

col_needed <- c("Q21","Q22","Q24_LOC_1","Q24_LOC_2","Q24_LOC_3","Q24_LOC_4")

check1 <- df1[,Q21_R := mean(Q21,na.rm = T)/mean(Q21_PAN,na.rm = T),by=region]

check1 works for one variable. I was looking for a solution where I can pass all needed variables and get the new variables calculated in a single line. So in this case something like passing col_needed. I tried below code as well -

check2 <- df1[,`:=`(paste0(col_needed,"_R"),(mean(col_needed,na.rm = T)/mean(paste0(col_needed,"_PAN"),na.rm = T))),by=region][]

However this gives me multiple warnings and the result is having all NAs. The warnings are - In mean(col_needed, na.rm = T) : argument is not numeric or logical: returning NA

Can you please suggest where I am going wrong.

user1412
  • 709
  • 1
  • 8
  • 25

1 Answers1

4

If I understand correctly, you could do the following:

df1[, paste(col_needed, "R", sep = "_") := 
      Map(function(x,y) mean(get(x), na.rm = TRUE)/mean(get(y), na.rm=TRUE), 
           col_needed, 
           paste(col_needed, "PAN", sep = "_")),
    by=region]
talat
  • 68,970
  • 21
  • 126
  • 157
  • Beat me to it. Alternatively, could use the `.SD` without needing `get`, but it's really just the same thing. Not suggesting a change, just showing a possible super minor tweak – Mike H. Apr 17 '18 at 12:37
  • @docendo, Thank you, this works. I understand by Map we are able to pass the function to each element col_needed. – user1412 Apr 17 '18 at 12:44
  • @MikeH.I had seen your post as well before you deleted. Thank you!! However I wanted to confirm with you on the usage of with = FALSE with Map. I am not clear why we had to use with = FALSE. Sorry if I am asking something very basic....still learning R and trying to get my basics clear – user1412 Apr 17 '18 at 12:47
  • @docendodiscimus Thank you for confirming !! – user1412 Apr 17 '18 at 12:48
  • @user1412, I used `with = FALSE` to select the columns by name. So `.SD[, col_needed, with = FALSE]` was a `data.table` of the columns we wanted – Mike H. Apr 17 '18 at 13:03