Slicing using dplyr based on condition in a column - r

Question

I have a dataframe containing repeated measurements for some individuals, the data.frame looks like this
# A tibble: 853 x 5
ID Test N_ind Pheno Week
<chr> <chr> <int> <dbl> <int>
1 02A01 Int 96 0 12
2 02A01 Int 96 0 24
3 02A01 Int 94 0 36
4 02A01 Int 90 0 48
7 02A01 Int 78 1 84
9 02A03 Int 96 0 12
10 02A03 Int 96 0 24
11 02A03 Int 94 0 36
19 02C03 Int 96 1 12
20 0202C03 Int 96 0 24
21 0202C03 Int 94 0 36
22 0202C03 Int 90 0 48
23 02E02 Int 96 0 12
24 02E02 Int 96 0 24
25 02E02 Int 94 1 36
26 02E02 Int 90 1 48
I want to subset the data.frame, first grouping by ID and then within the group selecting for those individuals with a 1 in the column Pheno the lowest Week value, but for those individuals with a 0 in the column Pheno the highest value in Week.
The optimal result should look like this:
ID Test N_ind Pheno Week
<chr> <chr> <int> <dbl> <int>
7 02A01 Int 78 1 84
11 02A03 Int 94 0 36
19 02C03 Int 96 1 12
22 0202C03 Int 90 0 48
25 02E02 Int 94 1 36
I have managed to do that for the 1 values but I am stuck in how to do it for the 0 values.
Here is my code:
df_sub <- data %>%
group_by(ID) %>%
arrange(Week, .by_group = TRUE) %>%
ungroup()
Any help would be much appreciated

Answer 1

You can try -
library(dplyr)
data %>%
group_by(ID) %>%
filter(Week == if(all(Pheno == 0)) max(Week) else min(Week[Pheno == 1])) %>%
ungroup
# ID Test N_ind Pheno Week
# <chr> <chr> <int> <int> <int>
#1 02A01 Int 78 1 84
#2 02A03 Int 94 0 36
#3 02C03 Int 96 1 12
#4 0202C03 Int 90 0 48
#5 02E02 Int 94 1 36
If all the Pheno values in a group are 0 return the max value or else from the Pheno = 1 values return the minimum one.

Answer 2

using dplyr and if, else in filter,
df %>%
group_by(ID) %>%
filter(if (Pheno == 0) Week == max(Week)
else Week == min(Week))
ID Test N_ind Pheno Week
<chr> <chr> <dbl> <dbl> <dbl>
1 02A01 Int 78 1 84
2 02A03 Int 94 0 36
3 02C03 Int 96 1 12
4 0202C03 Int 90 0 48
5 02E02 Int 90 1 48

Answer 3

We can first group_by ID, then filter with the max value in Pheno.
After that, we can filter with the conditions you want, separated by an OR | statement.
library(dplyr)
df %>% group_by(ID) %>%
filter(Pheno==max(Pheno))%>%
filter(Pheno==0 & Week==min(Week)|Pheno==1 & Week==max(Week))
# A tibble: 5 x 5
# Groups: ID [5]
ID Test N_ind Pheno Week
<chr> <chr> <int> <int> <int>
1 02A01 Int 78 1 84
2 02A03 Int 96 0 12
3 02C03 Int 96 1 12
4 0202C03 Int 96 0 24
5 02E02 Int 90 1 48

Answer 4

This is what I came up with:
data %>%
group_by(ID, Pheno) %>%
filter(case_when(
Pheno == 1 ~ Week == min(Week),
Pheno == 0 ~ Week == max(Week)
)) %>%
ungroup(Pheno) %>%
filter(case_when(
any(Pheno == 1) ~ Pheno == 1,
all(Pheno == 0) ~ Pheno == 0
)) %>%
ungroup()
Which results to:
# A tibble: 5 x 5
ID Test N_ind Pheno Week
<chr> <chr> <int> <int> <int>
1 02A01 Int 78 1 84
2 02A03 Int 94 0 36
3 02C03 Int 96 1 12
4 0202C03 Int 90 0 48
5 02E02 Int 94 1 36