R: subset data.frame based on column value using dplyr - r

Question

library(dplyr)
mydat1 <- data.frame(ID = c(1, 1, 2, 2),
Gender = c("Male", "Female", "Male", "Male"),
Score = c(30, 40, 20, 60))
mydat1 %>%
group_by(ID, Gender) %>%
slice(which.min(Score))
# A tibble: 3 x 3
# Groups: ID, Gender [3]
ID Gender Score
<dbl> <fctr> <dbl>
1 1 Female 40
2 1 Male 30
3 2 Male 20
I'm trying to group the rows by ID and Gender. And then I want to only keep the row with the lowest Score. The above code works perfectly because when ID == 2, I only kept the entry with the lower score.
mydat2 <- data.frame(ID = c(1, 1, 2, 2),
Gender = c("Male", "Female", "Male", "Male"),
Score = c(NA, NA, 20, 60))
mydat2 %>%
group_by(ID, Gender) %>%
slice(which.min(Score))
# A tibble: 1 x 3
# Groups: ID, Gender [1]
ID Gender Score
<dbl> <fctr> <dbl>
1 2 Male 20
However, when I have NAs, which.min doesn't work like I want it to because it'll not return a valid index. Instead, all of my ID == 1 entries are erased. My desired output in this scenario is:
# A tibble: 1 x 3
# Groups: ID, Gender [1]
ID Gender Score
<dbl> <fctr> <dbl>
1 1 Female NA
2 1 Male NA
1 2 Male 20
How can I modify my code to account for this?
Edit:
df2 <- structure(list(pubmed_id = c(23091106L, 23091106L), Gender = structure(c(4L,
4L), .Label = c("", "Both", "female", "Female", "Male"), class = "factor"),
Total_Carrier = c(NA, 1107)), class = c("grouped_df", "tbl_df",
"tbl", "data.frame"), row.names = c(NA, -2L), vars = "pubmed_id", drop = TRUE, indices = list(
0:1), group_sizes = 2L, biggest_group_size = 2L, labels = structure(list(
pubmed_id = 23091106L), class = "data.frame", row.names = c(NA,
-1L), vars = "pubmed_id", drop = TRUE, .Names = "pubmed_id"), .Names = c("pubmed_id",
"Gender", "Total_Carrier"))
> df2
# A tibble: 2 x 3
# Groups: pubmed_id [1]
pubmed_id Gender Total_Carrier
<int> <fctr> <dbl>
1 23091106 Female NA
2 23091106 Female 1107
In this example, I would want the desired output to only contain row 2 (i.e. the row with carrier sample size of 1107). However, I get the following result:
> df2 %>%
group_by(pubmed_id, Gender) %>%
slice(which.min(Total_Carrier) || 1)
# A tibble: 1 x 3
# Groups: pubmed_id, Gender [1]
pubmed_id Gender Total_Carrier
<int> <fctr> <dbl>
1 23091106 Female NA

Answer 1

which.min ignores the missing values, and returns integer(0) when the input vector contains solely NAs. You can add a condition check in the slice, i.e. when all Scores are NAs in a group, pick the first row:
mydat2 %>%
group_by(ID, Gender) %>%
slice({idx <- which.min(Score); if(length(idx) > 0) idx else 1})
# A tibble: 3 x 3
# Groups: ID, Gender [3]
# ID Gender Score
# <dbl> <fctr> <dbl>
#1 1 Female NA
#2 1 Male NA
#3 2 Male 20

Answer 2

You could also use arrange to sort your scores within your groups, and then slice to select the first row of each group. That way, if there are only NAs in the group, you would still select the first row:
mydat2 %>%
group_by(ID, Gender) %>%
arrange(ID,Gender,Score) %>%
slice(1)
ID Gender Score
<dbl> <fctr> <dbl>
1 1 Female NA
2 1 Male NA
3 2 Male 20

Answer 3

Here is another option with which and pmin
mydat2 %>%
group_by(ID, Gender) %>%
slice(pmin(1, which(Score == min(Score, na.rm = TRUE))[1], na.rm = TRUE))
# A tibble: 3 x 3
# Groups: ID, Gender [3]
# ID Gender Score
# <dbl> <fctr> <dbl>
#1 1 Female NA
#2 1 Male NA
#3 2 Male 20

Answer 4

A solution using data.table
library(data.table)
setDT(mydat2)
mydat2[, .(Score = sort(Score)[1]), by = .(ID, Gender)]
# ID Gender Score
# 1: 1 Male NA
# 2: 1 Female NA
# 3: 2 Male 20