Dividing data into quintiles using group_by - r

Question

I am looking for a way to change my way in such a way that it sorts the data into quintiles instead of the top 5 and bottom 5. My current code looks like this:
CombData <- CombData %>%
group_by(Date) %>%
mutate(
R=min_rank(Value),
E_P = case_when(
R < 6 ~ "5w",
R > max(R, na.rm =TRUE) - 5 ~ "5b",
TRUE ~ NA_character_)
) %>%
ungroup() %>%
arrange(Date, E_P)
My dataset is quite large therefore I will just provide sample data. The data I use is more complex and the code should, therefore, allow for varying lengths of the column Date and also for multiple values that are missing (NAs):
df <- data.frame( Date = c(rep("2010-01-31",16), rep("2010-02-28", 14)), Value=c(rep(c(1,2,3,4,5,6,7,8,9,NA,NA,NA,NA,NA,15),2))
Afterward, I would also like to test the minimum size of quintiles i.e. how many data points are minimum in each quintile in the entire dataset.
The expected output would look like this:
structure(list(Date = structure(c(14640, 14640, 14640, 14640,
14640, 14640, 14640, 14640, 14640, 14640, 14640, 14640, 14640,
14640, 14640, 14640, 14668, 14668, 14668, 14668, 14668, 14668,
14668, 14668, 14668, 14668, 14668, 14668, 14668, 14668), class = "Date"),
Value = c(1, 1, 2, 3, 4, 5, 6, 7, 8, 9, 15, NA, NA, NA, NA,
NA, 2, 3, 4, 5, 6, 7, 8, 9, 15, NA, NA, NA, NA, NA), R = c(1L,
1L, 3L, 4L, 5L, 6L, 7L, 8L, 9L, 10L, 11L, NA, NA, NA, NA,
NA, 1L, 2L, 3L, 4L, 5L, 6L, 7L, 8L, 9L, NA, NA, NA, NA, NA
), S_P = c("Worst", "Worst", "Worst", NA, NA, NA, NA, "Best",
"Best", "Best", NA, NA, NA, NA, NA, NA, "Worst", "Worst", NA, NA,
NA, NA, NA, "Best", "Best", NA, NA, NA, NA, NA)), row.names = c(NA,
-30L), class = c("tbl_df", "tbl", "data.frame"))

Answer 1

Probably, you could use something like this with quantile :
library(dplyr)
out <- CombData %>%
group_by(Date) %>%
mutate(S_P = case_when(Value <= quantile(Value, 0.2, na.rm = TRUE) ~ 'Worst',
Value >= quantile(Value, 0.8, na.rm = TRUE) ~ 'Best'))
You could change the value of quantile according to your preference.
To get minimum number of "Best" and "Worst" we can do :
out %>%
count(Date, S_P) %>%
na.omit() %>%
ungroup() %>%
select(-Date) %>%
group_by(S_P) %>%
top_n(-1, n)
# S_P n
# <chr> <int>
#1 Best 2
#2 Worst 2

Answer 2

When I understand you correctly, you want to rank your column 'Value' and mark those with rank below the quantile 20% with "worst" and those above 80% with "best". After that you want a table.
You could use use ave for both, the ranking and the quantile identification. The quantile function yields three groups, that you can identify with findInterval, code as a factor variable and label them at will. I'm not sure, though, which ranks should be included in the quantiles, I therefore make the E_P coding in two separate columns for comparison purposes.
dat2 <- within(dat, {
R <- ave(Value, Date, FUN=function(x) rank(x, na.last="keep"))
E_P <- ave(R, Date, FUN=function(x) {
findInterval(x, quantile(R, c(.2, .8), na.rm=TRUE))
})
E_P.fac <- factor(E_P, labels=c("worst", NA, "best"))
})
dat2 <- dat2[order(dat2$Date, dat2$E_P), ] ## order by date and E_P
Yields:
dat2
# Date Value E_P.fac E_P R
# 1 2010-01-31 1 worst 0 1.5
# 16 2010-01-31 1 worst 0 1.5
# 2 2010-01-31 2 <NA> 1 3.0
# 3 2010-01-31 3 <NA> 1 4.0
# 4 2010-01-31 4 <NA> 1 5.0
# 5 2010-01-31 5 <NA> 1 6.0
# 6 2010-01-31 6 <NA> 1 7.0
# 7 2010-01-31 7 <NA> 1 8.0
# 8 2010-01-31 8 best 2 9.0
# 9 2010-01-31 9 best 2 10.0
# 15 2010-01-31 15 best 2 11.0
# 10 2010-01-31 NA <NA> NA NA
# 11 2010-01-31 NA <NA> NA NA
# 12 2010-01-31 NA <NA> NA NA
# 13 2010-01-31 NA <NA> NA NA
# 14 2010-01-31 NA <NA> NA NA
# 17 2010-02-28 2 worst 0 1.0
# 18 2010-02-28 3 worst 0 2.0
# 19 2010-02-28 4 <NA> 1 3.0
# 20 2010-02-28 5 <NA> 1 4.0
# 21 2010-02-28 6 <NA> 1 5.0
# 22 2010-02-28 7 <NA> 1 6.0
# 23 2010-02-28 8 <NA> 1 7.0
# 24 2010-02-28 9 <NA> 1 8.0
# 30 2010-02-28 15 best 2 9.0
# 25 2010-02-28 NA <NA> NA NA
# 26 2010-02-28 NA <NA> NA NA
# 27 2010-02-28 NA <NA> NA NA
# 28 2010-02-28 NA <NA> NA NA
# 29 2010-02-28 NA <NA> NA NA
When I check the quintiles of the Rank column, it appears to be right.
quantile(dat2$R, c(.2, .8), na.rm=TRUE)
# 20% 80%
# 2.8 8.2
After that you could just make a table to get the numbers of each category.
with(dat2, table(Date, E_P.fac))
# E_P.fac
# Date worst <NA> best
# 2010-01-31 2 6 3
# 2010-02-28 2 6 1
Data
dat <- structure(list(Date = structure(c(1L, 1L, 1L, 1L, 1L, 1L, 1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 2L, 2L, 2L, 2L, 2L, 2L, 2L,
2L, 2L, 2L, 2L, 2L, 2L, 2L), .Label = c("2010-01-31", "2010-02-28"
), class = "factor"), Value = c(1, 2, 3, 4, 5, 6, 7, 8, 9, NA,
NA, NA, NA, NA, 15, 1, 2, 3, 4, 5, 6, 7, 8, 9, NA, NA, NA, NA,
NA, 15)), row.names = c(NA, -30L), class = "data.frame")