How to calculate a row-wise count of duplicates based on (element-wise) selected adjacent columns

Question

I have a data frame test:

group userID A_conf A_chall B_conf B_chall
1    220      1       1      1       2     
1    222      4       6      4       4     
2    223      6       5      3       2     
1    224      1       5      4       4    
2    228      4       4      4       4

The data contains responses per user (shown by userID) where each user may enter any value between 1 to 6 for both the measures:

conf
chall

They can also choose not to respond, resulting in an NA entry.

The test dataframe contains several columns like A, B, C, D and so on. Conf and Chall measures can be reported for each of these columns separately.

I am interested in making following comparisons:

A_conf & A_chall
B_conf & B_chall

IF any of these measures are equal, the Final counter should be incremented (as shown below).

group userID A_conf A_chall B_conf B_chall Final
1    220      1       1      1       2     1
1    222      4       6      4       4     1
2    223      6       5      3       2     0
1    224      1       5      4       4     1
2    228      4       4      4       4     2

I am struggling with the Final counter. What script would help me achieve this functionality?

For reference, the dput of the test dataframe set is shared below:

dput(test):

structure(list(group = c(1L, 1L, 2L, 1L, 2L),

userID = c(220L, 222L, 223L, 224L, 228L),

A_conf = c(1L, 4L, 6L, 1L, 4L),

A_chall = c(1L, 6L, 5L, 5L, 4L),

B_conf = c(1L, 4L, 3L, 4L, 4L),

B_chall = c(2L, 4L, 2L, 4L, 4L)),

class = "data.frame", row.names = c(NA, -5L))

I tried a code like this:

test$Final = as.integer(0)   # add a column to keep counts
count_inc = as.integer(0)    # counter variable to increment in steps of 1

for (i in 1:nrow(test)) {

    count_inc = 0

    if(!is.na(test$A_conf[i] == test$A_chall[i]))
    {
      count_inc = 1
      test$Final[i] = count_inc
    }#if

    else if(!is.na(test$A_conf[i] != test$A_chall[i]))
    {
      count_inc = 0
      test$Final[i] = count_inc
    }#else if
}#for

The above code has been written to work ONLY on the columns A_conf and A_chall. The problem is, it fills the Final column with all 1's whether the entered values (by users) are equal or not.

score 4 · Accepted Answer · answered Dec 20 '18 at 11:33

A base R solution assuming you have equal number of "conf" and "chall" columns

#Find indexes of "conf" column
conf_col <- grep("conf", names(test))

#Find indexes of "chall" column
chall_col <- grep("chall", names(test))

#compare element wise and take row wise sum
test$Final <- rowSums(test[conf_col] == test[chall_col])


test
#  group userID A_conf A_chall B_conf B_chall Final
#1     1    220      1       1      1       2     1
#2     1    222      4       6      4       4     1
#3     2    223      6       5      3       2     0
#4     1    224      1       5      4       4     1
#5     2    228      4       4      4       4     2

Can also be done in one-liner

rowSums(test[grep("conf", names(test))] == test[grep("chall", names(test))])

Many thanks, the suggested solution is easy to follow. Just a quick question though, how to deal with NAs. In my original question, I forgot to mention that some of the entries are missing and contain NAs. — Sandy, Dec 20 '18 at 11:44
@Sadiaz `rowSums` has got `na.rm` parameter., you can set it to `TRUE` like `rowSums(test[conf_col] == test[chall_col], na.rm = TRUE)` — Ronak Shah, Dec 20 '18 at 11:46

score 2 · Answer 2 · answered Dec 20 '18 at 11:22

With tidyverse you can do:

df %>%
 select(-Final) %>%
 rowid_to_column() %>% #Creating an unique row ID
 gather(var, val, -c(group, userID, rowid)) %>% #Reshaping the data
 arrange(rowid, var) %>% #Arranging by row ID and by variables
 group_by(rowid) %>% #Grouping by row ID
 mutate(temp = gl(n()/2, 2)) %>% #Creating a grouping variable for different "_chall" and "_conf" variables
 group_by(rowid, temp) %>% #Grouping by row ID and the new grouping variables
 mutate(res = ifelse(val == lag(val), 1, 0)) %>% #Comparing whether the different "_chall" and "_conf" have the same value
 group_by(rowid) %>% #Grouping by row ID
 mutate(res = sum(res, na.rm = TRUE)) %>% #Summing the occurrences of "_chall" and "_conf" being the same
 select(-temp) %>% 
 spread(var, val) %>% #Returning the data to its original form
 ungroup() %>%
 select(-rowid)

  group userID   res A_chall A_conf B_chall B_conf
  <int>  <int> <dbl>   <int>  <int>   <int>  <int>
1     1    220    1.       1      1       2      1
2     1    222    1.       6      4       4      4
3     2    223    0.       5      6       2      3
4     1    224    1.       5      1       4      4
5     2    228    2.       4      4       4      4

score 2 · Answer 3 · answered Dec 20 '18 at 11:29

You can try this tidyverse as well. Some less lines compared to the other answer ;)

library(tidyverse)
d %>% 
  as.tibble() %>% 
  gather(k, v, -group,-userID) %>% 
  separate(k, into = c("letters", "test")) %>% 
  spread(test, v) %>% 
  group_by(userID) %>% 
  mutate(final = sum(chall == conf)) %>% 
  distinct(userID, final) %>% 
  ungroup() %>% 
  right_join(d)
# A tibble: 5 x 7
  userID final group A_conf A_chall B_conf B_chall
   <int> <int> <int>  <int>   <int>  <int>   <int>
1    220     1     1      1       1      1       2
2    222     1     1      4       6      4       4
3    223     0     2      6       5      3       2
4    224     1     1      1       5      4       4
5    228     2     2      4       4      4       4

How to calculate a row-wise count of duplicates based on (element-wise) selected adjacent columns

3 Answers3