如何使 mutate(map()) 识别组？

Question

db = tibble(
x = runif(1000, 1, 10),
t = rpois(1000, 5),
Group = rpois(1000, 5)
)

for (i in c(1:1000) {
  db$l[i] <- mean(db$x[(db$t < db$t[i]) & (db$group == db$group[i])])
  }

你如何运行这个for更快？ mutate() 和 map() 的组合应该更快，但是

f <- function(lim) mean(a$x[a$t < lim])

db %>% group_by(Group) %>%
 mutate(l = map_dbl(t, function(y) mean(db$x[db$t < y])))

不承认 db 会是 grouped by = Group。

Answer 1

你可以试试

set.seed(20211212)

library(dplyr)
library(purrr)

db %>% 
  group_by(group) %>% 
  mutate(l = map_dbl(t, ~mean(db$x[db$t < .x])))

这个returns

# A tibble: 1,000 x 4
# Groups:   group [14]
       x     t group     l
   <dbl> <int> <int> <dbl>
 1  2.53     6     6  5.48
 2  1.23     4     8  5.37
 3  4.51     3     3  5.40
 4  2.45     8     3  5.49
 5  7.23     6     7  5.48
 6  8.11     5     5  5.35
 7  2.14     4     1  5.37
 8  3.17     4     4  5.37
 9  2.69     3     7  5.40
10  7.85     5     5  5.35
# ... with 990 more rows

但在你的情况下，我相信你想要

db %>% 
  group_by(group) %>% 
  mutate(l = map_dbl(t, ~mean(x[t < .x])))

哪个returns

# A tibble: 1,000 x 4
# Groups:   group [14]
       x     t group     l
   <dbl> <int> <int> <dbl>
 1  2.53     6     6  5.85
 2  1.23     4     8  4.64
 3  4.51     3     3  4.23
 4  2.45     8     3  5.19
 5  7.23     6     7  5.44
 6  8.11     5     5  5.77
 7  2.14     4     1  5.52
 8  3.17     4     4  4.69
 9  2.69     3     7  5.73
10  7.85     5     5  5.77
# ... with 990 more rows

如何使 mutate(map()) 识别组？

How to make a mutate(map()) recognize groups?

r

dplyr

purrr