Sum NA across specific columns in R

Question

I have data such as this:

  data_in <- read_table2("Id    Q62_1   Q62_2   Q3_1    Q3_2    Q3_3    Q3_4    Q3_5
1   Yes Sometimes                   
2       Always                  
3                           
4   No  Always  Yes             
5                           
6       Always  No  Likely  Yes Always  Always
7   Yes Sometimes   Maybe   Unlikely        Sometimes   Sometimes
8       Always  Yes Likely  No  Always  Always
9       Sometimes       Unlikely        Sometimes   Sometimes
10  No      No  Likely  Maybe       
11      Sometimes   Maybe   Unlikely        Sometimes   Sometimes
12      Always  Yes Likely      Always  Always
")

I would like to calculate the number of missing response within columns that start with Q62 and then from columns Q3_1 to Q3_5 separately.

I know that rowSums is handy to sum numeric variables, but is there a dplyr/piped equivalent to sum na's?

For example, if this were numeric data and I wanted to sum the q62 series, I could use the following:

data_in %>% 
  mutate(Q62_NA = rowSums(select(.,"Q62_1", "Q62_2"))

But how do I sum NAs?

My output should look something like this:

data_out <- read_table2("Id Q62_1   Q62_2   Q3_1    Q3_2    Q3_3    Q3_4    Q3_5    Q62_NA  Q3_NA
1   Yes Sometimes                       0   5
2       Always                      1   5
3                               2   5
4   No  Always  Yes                 0   5
5                               2   5
6       Always  No  Likely  Yes Always  Always  1   
7   Yes Sometimes   Maybe   Unlikely        Sometimes   Sometimes   0   1
8       Always  Yes Likely  No  Always  Always  1   0
9       Sometimes       Unlikely        Sometimes   Sometimes   1   1
10  No      No  Likely  Maybe           1   2
11      Sometimes   Maybe   Unlikely        Sometimes   Sometimes   1   1
12      Always  Yes Likely      Always  Always  1   1
")

Thank you!!

Answer 1

We can wrap the select with is.na to convert it to a logical matrix and then do the rowSums on that matrix to sum the number of TRUE elements per row

library(dplyr)
data_in %>% 
   mutate(Q62_NA = rowSums(is.na(select(.,"Q62_1", "Q62_2"))))

---

Or an option with c_across and rowwise

 data_in %>%
    rowwise %>%
    mutate(Q62_NA = sum(is.na(c_across(starts_with('Q6')))))

Answer 2

Here is a base R option

transform(
  data_in,
  Q62_NA = rowSums(is.na(data_in[grepl("Q62",names(data_in))])),
  Q3_NA = rowSums(is.na(data_in[grepl("Q3",names(data_in))]))
)

which gives

   Id     Q62_1     Q62_2      Q3_1      Q3_2      Q3_3      Q3_4 Q3_5 Q62_NA
1   1       Yes Sometimes      <NA>      <NA>      <NA>      <NA> <NA>      0
2   2    Always      <NA>      <NA>      <NA>      <NA>      <NA> <NA>      1
3   3      <NA>      <NA>      <NA>      <NA>      <NA>      <NA> <NA>      2
4   4        No    Always       Yes      <NA>      <NA>      <NA> <NA>      0
5   5      <NA>      <NA>      <NA>      <NA>      <NA>      <NA> <NA>      2
6   6    Always        No    Likely       Yes    Always    Always <NA>      0
7   7       Yes Sometimes     Maybe  Unlikely Sometimes Sometimes <NA>      0
8   8    Always       Yes    Likely        No    Always    Always <NA>      0
9   9 Sometimes  Unlikely Sometimes Sometimes      <NA>      <NA> <NA>      0
10 10        No        No    Likely     Maybe      <NA>      <NA> <NA>      0
11 11 Sometimes     Maybe  Unlikely Sometimes Sometimes      <NA> <NA>      0
12 12    Always       Yes    Likely    Always    Always      <NA> <NA>      0
   Q3_NA
1      5
2      5
3      5
4      4
5      5
6      1
7      1
8      1
9      3
10     3
11     2
12     2