Applying custom ranking values within groups using dplyr

[gjgetzinger]

I would like to use dplyr::summarize with a custom function that operates on groups within a data frame. Specifically, I want to create a custom rank based on the relative score of grouped entries. See example below:

library(dplyr)

dat <- data.frame(grp = rep(LETTERS[1:3], each = 3),
                  score = rnorm(n = 9), 
                  correct = rep(c(T,F,F), 3)) 
dat 
#>   grp      score correct
#> 1   A -0.4197577    TRUE
#> 2   A -1.8704389   FALSE
#> 3   A  0.9077843   FALSE
#> 4   B  0.7294849    TRUE
#> 5   B -1.5251637   FALSE
#> 6   B -0.3363766   FALSE
#> 7   C -0.4401990    TRUE
#> 8   C  1.5558136   FALSE
#> 9   C -0.8925630   FALSE

# for each correct entry, determine the relative rank position (RRP) within the group...
# RRP = 0.5 * (1 - ((BC - WC)/(TC - 1)))
# BC : candidates with better score
# WC: candidates with worse score 
# TC: total candidates 

# in the case of grp == 'A'...
with(dat %>%
       group_by(grp) %>%
       filter(grp == 'A'), {
         0.5 * (1 - (length(which(score > score[correct])) - length(which(score < score[correct]))) /
                  (length(score) - 1))
       }) 
#> [1] 0.5

# work around using sapply...
sapply(
  unique(dat$grp),
  function(x_grp){
    with(dat %>%
           group_by(grp) %>%
           filter(grp == !!x_grp), {
             0.5 * (1 - (length(which(score > score[correct])) - length(which(score < score[correct]))) /
                      (length(score) - 1))
           }) 
  }) %>% cbind(dat[which(dat$correct),], rrp = .)
#>   grp      score correct rrp
#> 1   A -0.4197577    TRUE 0.5
#> 4   B  0.7294849    TRUE 1.0
#> 7   C -0.4401990    TRUE 0.5

# what I want to do... 
dat %>% group_by(grp) %>% summarize(rrp = funs())

^{Created on 2019-07-23 by the reprex package (v0.3.0)}

[pieterjanvc]

Hi,

Would you clarify what exactly it is you like, because it seems your code is already producing the result you're asking for ...

Thanks

[gjgetzinger]

Thanks for your response. I updated the question to clarify. Mostly, I am interested in a summarize function that will calculate a value based on other values in a group.