dplyr::mutate (assign na.rm =TRUE)

Question

I have a data.frame that has 100 variables. I want to get the sum of three variables only using mutate (not summarise).

If there is NA in a

Answer 1

Where better = tidyr:

df %>%
    mutate(rn = row_number()) %>%
    gather(var, varNum, var1:var3) %>%
    group_by(rn) %>%
    mutate(sum = sum(varNum, na.rm = TRUE)) %>% 
    spread(var, varNum)

In case your dataset is poised to grow...

Answer 2

rowwise() is my go-to function. It's like group_by() but it treats each row as an individual group.

df %>% rowwise() %>% mutate(Sum = sum(c(var1, var2, var3), na.rm = TRUE))

Answer 3

We can use Reduce with +

df %>% 
     mutate_each(funs(replace(., is.na(.), 0)), var1:var3) %>% 
     mutate(Sum = Reduce(`+`, .))      
#   var1 var2 var3 Sum
#1    4    5    0   9
#2    2    0    3   5
#3    1    2    4   7
#4    0    3    5   8
#5    3    0    2   5
#6    1    1    5   7

---

Or with rowSums

df %>% 
   mutate(Sum = rowSums(.[names(.)[1:3]], na.rm = TRUE))
#   var1 var2 var3 Sum
#1    4    5   NA   9
#2    2   NA    3   5
#3    1    2    4   7
#4   NA    3    5   8
#5    3   NA    2   5
#6    1    1    5   7

Benchmarks

set.seed(24)
df1 <- as.data.frame(matrix(sample(c(NA, 1:5), 1e6 *3, replace=TRUE),
                dimnames = list(NULL, paste0("var", 1:3)), ncol=3))
system.time({
df1 %>% rowwise() %>% mutate(Sum = sum(c(var1, var2, var3), na.rm = TRUE))
})
# user  system elapsed 
#  21.50    0.03   21.66 

system.time({
df1 %>%
    mutate(rn = row_number()) %>%
    gather(var, varNum, var1:var3) %>%
    group_by(rn) %>%
    mutate(sum = sum(varNum, na.rm = TRUE)) %>% 
    spread(var, varNum)})
 # user  system elapsed 
 #  5.96    0.39    6.37 


system.time({
replace(df1, is.na(df1), 0) %>% mutate(sum = var1 + var2 + var3)
})

# user  system elapsed 
#   0.17    0.01    0.19 

system.time({
df1 %>% 
     mutate_each(funs(replace(., is.na(.), 0)), var1:var3) %>% 
     mutate(Sum = Reduce(`+`, .))      
})
# user  system elapsed 
#   0.10    0.02    0.11 

system.time({
df1 %>% 
   mutate(Sum = rowSums(.[names(.)[1:3]], na.rm = TRUE))
   })
# user  system elapsed 
#   0.04    0.00    0.03