Nesting tibbles and performing calculations on each with a group_by - r

Question

I want to find out what the dominant class of of each id_b is. To calculate it I need to find out the sum of size per class for each id_b. Whichever class is largest is the new dominant class assigned to id_b.
The script below does what I want it to do but it feels quite clunky and over complicated. I've not worked with nested data much before so I'm not sure if I've used the best methods possible Can anyone think of a neater way of achieving the same output in tidyverse or data.table?
Thanks!
library(tidyverse)
# sample data
set.seed(123)
input <- tibble(id_a = c(letters[seq(1,10)]),
size = runif(10, min = 10, max = 50),
class = c("x","x","y","x","y",
"y","x","y","x","x"),
id_b = c("A1","A1","B1","B1","B1",
"C1","C1","C1","D1","E1"))
print(input)
id_a size class id_b
<chr> <dbl> <chr> <chr>
1 a 23.6 x A1
2 b 43.6 x A1
3 c 23.9 y B1
4 d 23.4 x B1
5 e 29.1 y B1
6 f 45.7 y C1
7 g 44.6 x C1
8 h 25.6 y C1
9 i 41.1 x D1
10 j 48.4 x E1
# nest input to create a nested tibble for each id_b
input_nest <- input %>% group_by(id_b) %>% nest()
# calculate dominant class
input_nest_dominant <- input_nest %>% mutate(DOMINANT_CLASS = lapply(data, function(x){
# group each nested tibble by class, and calculate total size. Then find the biggest size and extract
# the class value
output <- x %>% group_by(class) %>%
summarise(total_size = sum(size)) %>%
top_n(total_size, n = 1) %>%
pull(class)
return(output)
} ))
# unnest to end up with a tibble
input_nest_dominant_clean <- input_nest_dominant %>%
unnest(cols = c(DOMINANT_CLASS)) %>%
select(-data) %>%
ungroup()
print(input_nest_dominant_clean)
id_b DOMINANT_CLASS
<chr> <chr>
1 A1 x
2 B1 y
3 C1 y
4 D1 x
5 E1 x

Answer 1

From this example, you don't need nest at all, just calculate it using group_by and summarize.
input %>%
group_by(id_b, class) %>%
summarize(size = sum(size)) %>%
group_by(id_b) %>%
summarize(DOMINANT_CLASS = class[which.max(size)])
#> # A tibble: 5 x 2
#> id_b DOMINANT_CLASS
#> <chr> <chr>
#> 1 A1 x
#> 2 B1 y
#> 3 C1 y
#> 4 D1 x
#> 5 E1 x

Answer 2

Here is a base R solution, which used aggregate twice, i.e.,
agg <-aggregate(size ~ class + id_b, input, FUN = sum)
output <- aggregate(agg[-2],agg[2],FUN = max)[-3]
or a more compact version
output <- aggregate(.~id_b,
aggregate(size ~ class + id_b,
input,
FUN = function(v) sum(v)),
FUN = function(v) tail(sort(v),1))[-3]
such that
> output
id_b class
1 A1 x
2 B1 y
3 C1 y
4 D1 x
5 E1 x

Answer 3

You can do just 1 sort, and remove all duplicates. Something like:
input %>% arrange(desc(size)) %>% filter(!duplicated(id_b)) %>% arrange(id_b)
# A tibble: 5 x 4
id_a size class id_b
<chr> <dbl> <chr> <chr>
1 b 41.5 x A1
2 e 47.6 y B1
3 h 45.7 y C1
4 i 32.1 x D1
5 j 28.3 x E1
If the order of id_b is not important you can omit the last arrange
Or in base R:
input = input[order(-input$size),]
input[!duplicated(input$id_b),]