Count the number of non-NA numeric values of each row in dplyr

Question

I create a dataframe df.

df <- data.frame (id = 1:10, 
    var1 = 10:19,
    var2 = sample(c(1:2,NA), 10, replace=T),
    var3 = sample(c(3:5, NA), 10, replace=T))

What I need is a new column var4, which count the number of non-NA values of each row (excluding the id column). So for example, if a row is like var1=19, var2=1, var3=NA, then var4=2. I could not find a good way to do this in dplyr. something like:

df %in% mutate(var4= ... )

I appreciate if anyone can help me with that.

Answer 1

Use select + is.na + rowSums, select(., -id) returns the original data frame (.) with id excluded, and then count number of non-NA values with rowSums(!is.na(...)):

df %>% mutate(var4 = rowSums(!is.na(select(., -id))))

#   id var1 var2 var3 var4
#1   1   10   NA    4    2
#2   2   11    1   NA    2
#3   3   12    2    5    3
#4   4   13    2   NA    2
#5   5   14    1   NA    2
#6   6   15    1   NA    2
#7   7   16    1    5    3
#8   8   17   NA    4    2
#9   9   18   NA    4    2
#10 10   19   NA   NA    1

Answer 2

I know the OP asked for a dplyr solution, but base R is straightforward here:

df$var4 <- rowSums(!is.na(df[,2:4]))

rowSums calculates the number of values that are not NA (!is.na) in columns 2 - 4.

Note, this is summing the logical vector generated by is.na, which is distinct from:

rowSums(df[,2:4], na.rm = TRUE)

Which drops the NAs and then sums the remaining values.

Answer 3

Another solution using only base-r

data.frame(df, var4 = apply(df[,-1], 1, function(x) sum(!is.na(x))))

   id var1 var2 var3 var4
1   1   10    1    5    3
2   2   11    2    5    3
3   3   12    2    5    3
4   4   13   NA    3    2
5   5   14   NA    5    2
6   6   15    1    5    3
7   7   16   NA    3    2
8   8   17   NA    4    2
9   9   18   NA    3    2
10 10   19    1    4    3