R group by, counting non-NA values

Question

I have a dataframe that has a scattering of NA's

toy_df
# Y  X1 X2 Label
# 5  3  3  A
# 3  NA 2  B
# 3  NA NA C
# 2  NA 6  B

I want to group this by the label field, and count how many non NA values are in each variable for each label.

desired output:
# Label Y  X1 X2
# A     1  1  1
# B     2  0  2
# C     1  0  0

I've done this using loops at the moment, but it's slow and untidy and I'm sure there's a better way.

Aggregate seems to get half way there, but it includes NA's in the count.

aggregate(toy_df, list(toy_df$label), FUN=length)

Any ideas appreciated...

Answer 1

We can use data.table. Convert the 'data.frame' to 'data.table' (setDT(toy_df)), grouped by 'Label', loop through the Subset of Data.table (.SD) and get the sum of non-NA values (!is.na(x))

library(data.table)
setDT(toy_df)[, lapply(.SD, function(x) sum(!is.na(x))), by = Label]
#   Label Y X1 X2
#1:     A 1  1  1
#2:     B 2  0  2
#3:     C 1  0  0

---

Or with dplyr using the same methodology

library(dplyr)
toy_df %>% 
      group_by(Label) %>%
      summarise_each(funs(sum(!is.na(.))))

---

Or a base R option with by and colSums grouped by the 4th column on logical matrix (!is.na(toy_df[-4]))

by(!is.na(toy_df[-4]), toy_df[4], FUN = colSums)

Or with rowsum with similar approach as in by except using the rowsum function.

rowsum(+(!is.na(toy_df[-4])), group=toy_df[,4])
#  Y X1 X2
#A 1  1  1
#B 2  0  2
#C 1  0  0

Answer 2

aggregate(cbind(toy_df$Y, toy_df$X1, toy_df$X2), list(toy_df$label),
          FUN = function (x) sum(!is.na(x)))

Answer 3

Or in base R

aggregate(toy_df[,1:3], by=list(toy_df$Label), FUN=function(x) { sum(!is.na(x))})