set.seed(2024)
df <- data.frame(x1=rep(c(1:4),each=12), x2=rep(c("A","B","C"),each=4,times=4), y=runif(48), z=rep(c(12,4,19,8,58,2,33,32,25,70,5,25),each=4))

I’m trying to select rows for all unique letters that have the highest value in z.
In my example, A would be the rows:

37  4  A 0.0541251726 70
38  4  A 0.9661198864 70
39  4  A 0.0221378489 70
40  4  A 0.6473674888 70

for B:

17  2  B 0.1103110805 58
18  2  B 0.8632152062 58
19  2  B 0.1310691827 58
20  2  B 0.2954267922 58

For C, the largest z value is 25 when x1==3 and x1==4, so select the rows with the lowest value in x1:

33  3  C 0.1762014020 25
34  3  C 0.9733794478 25
35  3  C 0.6751627198 25
36  3  C 0.6809547509 25

Together, this is the expected result:

   x1 x2            y  z
37  4  A 0.0541251726 70
38  4  A 0.9661198864 70
39  4  A 0.0221378489 70
40  4  A 0.6473674888 70
17  2  B 0.1103110805 58
18  2  B 0.8632152062 58
19  2  B 0.1310691827 58
20  2  B 0.2954267922 58
33  3  C 0.1762014020 25
34  3  C 0.9733794478 25
35  3  C 0.6751627198 25
36  3  C 0.6809547509 25

This almost does the job:

library(dplyr)
df %>% group_by(x2) %>% slice(which.max(z))

but I need to select all other rows belonging to the same x1.