I have the following table:
# Inputs
require(dplyr)
set.seed(123)
df <- tibble(g = rep(x = c("A", "B", "C"), times = c(3, 5, 7)),
a = sample(x = 1:100, size = 15),
b = sample(x = 1:100, size = 15))
df
# A tibble: 15 × 3
g a b
<chr> <int> <int>
1 A 87 8
2 A 35 51
3 A 40 74
4 B 30 50
5 B 12 98
6 B 31 86
7 B 97 76
8 B 64 84
9 C 14 46
10 C 71 17
11 C 67 62
12 C 23 92
13 C 79 54
14 C 85 35
15 C 37 79
I also define a function (for this example, very simple) that takes a whole data frame and returns a single numeric value:
# Function
myFUN <- function(x){
mean(x$a + x$b)
}
What I am looking for is to apply the group_by and summarise functions to get a table with the results for each group.
I thought it could be as simple as doing the following:
# What I got (incorrect results)
df %>%
group_by(g) %>%
summarise(res = myFUN(.))
# A tibble: 3 × 2
g res
<chr> <dbl>
1 A 112.
2 B 112.
3 C 112.
But as you can see, all the result (res) values are the same, because . refers to the whole initial table and not to the subset tables withing each group.
I leave an example of the expected result using a loop:
# Expected
out <- list()
for(i in unique(df$g)){
out[[length(out) + 1]] <- tibble(g = i,
res = df %>% filter(g == i) %>% myFUN)
}
out |> bind_rows()
# A tibble: 3 × 2
g res
<chr> <dbl>
1 A 98.3
2 B 126.
3 C 109.
Solution (taken from the answer of @Limey)
df %>%
group_by(g) %>%
group_map(.f = function(.x, .y){
.x %>%
summarise(res = myFUN(.)) %>%
mutate(g = .y$g, .before = 1)
}) %>%
bind_rows()
I don't get your expected results, but I think this gives you what you want. .x takes the place of your . because group_map expects a function with two arguments.
df %>%
group_by(g) %>%
group_map(
function(.x, .y) {
.x %>%
summarise(res = mean(a + b)) %>%
add_column(g = .y$g, .before = 1)
}
) %>% bind_rows()
# A tibble: 3 × 2
g res
<chr> <dbl>
1 A 78.7
2 B 101.
3 C 117.
And, of course, this can be reduced to
df %>%
group_by(g) %>%
summarise(res = mean(a + b))
(with identical results) in this case, but I accept you have simplified your real use-case to produce your MWE.