问题
Sample data frame data
uid bas_id dist2mouth type
2020 2019 W3A9101601 2.413629 1
2021 2020 W3A9101601 2.413629 1
2022 2021 W3A9101602 2.413629 1
2023 2022 W3A9101602 3.313893 1
2032 2031 W3A9101602 3.313893 1
2033 2032 W3A9101602 3.313893 1
2034 2033 W3A9101602 3.313893 1
15023 15022 W3A9101601 1.349000 2
15025 15024 W3A9101601 3.880000 2
15026 15025 W3A9101602 3.880000 2
15027 15026 W3A9101602 0.541101 2
16106 17097 W3A9101602 1.349000 2
For each row I'd like to calculate how many rows of type=2 within the same bas_id have a lower dist2mouth. Effectively how many rows type=2 are downstream of each row. Store it as ds_n_type2. So far I've tried dplyr
ds <- data %>%
group_by(id) %>%
summarize(n_ds = sum(dist2mouth > id[dist2mouth]))
I would then like to find the closest row type=2 to each row type=1 within the same bas_id maybe using which in a for or apply loop. Store it as closest_uid_type2. Maybe something like
which(abs(x[i:n]-x[i])==min(abs(x[i:n]-x[i])))
Happy to clarify
Edit 2 Desired output amended
uid bas_id dist2mouth type ds_n_type2 closest_uid_type2
2020 2019 W3A9101601 2.413629 1 1 15022
2021 2020 W3A9101601 2.413629 1 1 15022
2022 2021 W3A9101602 2.413629 1 2 15022
2023 2022 W3A9101602 3.313893 1 2 15024
2032 2031 W3A9101602 3.313893 1 2 15024
2033 2032 W3A9101602 3.313893 1 2 15024
2034 2033 W3A9101602 3.313893 1 2 15024
15023 15022 W3A9101601 1.349000 2 - -
15025 15024 W3A9101601 3.880000 2 - -
15026 15025 W3A9101602 3.880000 2 - -
15027 15026 W3A9101602 0.541101 2 - -
17097 W3A9101602 1.349000 2 - -
回答1:
Try this:
require(dplyr)
df %>%
group_by(bas_id) %>%
mutate(n_ds = match(dist2mouth,sort(dist2mouth))-1) %>%
mutate(closest_uid=apply(
sapply(dist2mouth,function(i)abs(i-dist2mouth)),
2,function(n) uid[which(n==sort(n)[2])])) %>%
data.frame()
Output:
uid dist2mouth bas_id type n_ds closest_uid
1 1 10 1 1 2 4
2 2 5 1 2 0 3
3 3 6 1 1 1 2
4 4 11 1 1 3 1
5 5 3 2 2 0 6
6 6 4 2 1 1 5
Edit:
This may not be the most elegant, but here's one way to solve the updated question (pending time to refine it):
df$ds_n_type2[df$type==1] <- sapply(as.numeric(row.names(df[df$type==1,])),
function(x) sum(as.numeric(df$dist2mouth[x]) > as.numeric(df$dist2mouth[df$bas_id==df$bas_id[x] & df$type==2])))
df$closest_uid_type2[df$type==1] <- sapply(as.numeric(row.names(df[df$type==1,])),
function(x) df$uid[which(df$dist2mouth==df$dist2mouth[df$bas_id==df$bas_id[x] & df$type==2][which.min(abs(c(df$dist2mouth[df$bas_id==df$bas_id[x] & df$type==2])-df$dist2mouth[x]))])[1]])
Output:
uid bas_id dist2mouth type ds_n_type2 closest_uid_type2
1: 2019 W3A9101601 2.413629 1 1 15022
2: 2020 W3A9101601 2.413629 1 1 15022
3: 2021 W3A9101602 2.413629 1 2 15022
4: 2022 W3A9101602 3.313893 1 2 15024
5: 2031 W3A9101602 3.313893 1 2 15024
6: 2032 W3A9101602 3.313893 1 2 15024
7: 2033 W3A9101602 3.313893 1 2 15024
8: 15022 W3A9101601 1.349000 2 NA NA
9: 15024 W3A9101601 3.880000 2 NA NA
10: 15025 W3A9101602 3.880000 2 NA NA
11: 15026 W3A9101602 0.541101 2 NA NA
12: 17097 W3A9101602 1.349000 2 NA NA
回答2:
I found it easier to split your data frame and use purrr:map
library(purrr)
L <- map(split(df, df$bas_id), ~split(.x, .x$type))
# $`1`
# $`1`$`1`
# uid dist2mouth bas_id type
# 1: 1 10 1 1
# 2: 3 6 1 1
# 3: 4 11 1 1
# $`1`$`2`
# uid dist2mouth bas_id type
# 1: 2 5 1 2
# $`2`
# $`2`$`1`
# uid dist2mouth bas_id type
# 1: 6 4 2 1
# $`2`$`2`
# uid dist2mouth bas_id type
# 1: 5 3 2 2
Answer to first Q
twolessthanone <- map_dbl(L, ~sum(.x$'2'$dist2mouth < .x$'1'$dist2mouth))
# 1 2
# 3 1
This is a named vector, don't let the extra numbers confuse you
str(twolessthanone)
# Named num [1:2] 3 1
# - attr(*, "names")= chr [1:2] "1" "2"
Answer to 2nd Q
nearestonetotwo <- map(L, ~.x$'1'[which.min(abs(.x$'2'$dist2mouth - .x$'1'$dist2mouth)),])
# $`1`
# uid dist2mouth bas_id type
# 1: 3 6 1 1
# $`2`
# uid dist2mouth bas_id type
# 1: 6 4 2 1
来源：https://stackoverflow.com/questions/46242567/count-values-less-than-x-and-find-nearest-values-to-x-by-multiple-groups