calculate median by groups with creating categorical variable in R - r

Question

Here example of my data
dput(mydat)
structure(list(ID.group = c(NA, 10150591L, NA, 10150591L, NA,
10150591L, NA, 68837296L, NA, 68837296L, NA, 68837296L, NA, 124771228L,
NA, 124771228L), UserID = c(NA, 181078814L, NA, 88578209L, NA,
30240768L, NA, 334686951L, NA, 297170412L, NA, 265332359L, NA,
216632504L, NA, 5272133L), countlike = c(NA, 44L, NA, 50L, NA,
99L, NA, 1L, NA, 1L, NA, 15L, NA, 41L, NA, 20L), statistics.snt = structure(c(1L,
2L, 1L, 2L, 1L, 2L, 1L, 2L, 1L, 2L, 1L, 2L, 1L, 2L, 1L, 2L), .Label = c("",
"fb"), class = "factor"), statistics.created_at = structure(c(1L,
8L, 1L, 4L, 1L, 7L, 1L, 2L, 1L, 2L, 1L, 5L, 1L, 3L, 1L, 6L), .Label = c("",
"10.04.2020 9:14", "11.04.2020 0:01", "11.04.2020 19:22", "12.04.2020 19:45",
"12.04.2020 6:54", "13.04.2020 20:47", "17.04.2020 23:02"), class = "factor"),
statistics.updated_at = structure(c(1L, 8L, 1L, 7L, 1L, 6L,
1L, 3L, 1L, 3L, 1L, 4L, 1L, 5L, 1L, 2L), .Label = c("", "22.04.2020 12:27",
"22.04.2020 12:51", "22.04.2020 14:19", "22.04.2020 5:41",
"22.04.2020 6:18", "22.04.2020 7:37", "30.04.2020 16:55"), class = "factor"),
statistics.is_recount = structure(c(1L, 2L, 1L, 2L, 1L, 2L,
1L, 2L, 1L, 2L, 1L, 2L, 1L, 2L, 1L, 2L), .Label = c("", "False"
), class = "factor")), class = "data.frame", row.names = c(NA,
-16L))
I want calculate the median for countlike by ID group
library(psych)
describeBy(mydat,mydat$ID.group)
but i didn't get needed result, i get all descriptive statistics.
How can i get results like
ID group median countlike
10150591 50
68837296 1
Then how calculate categorical variable for UserID?
For example. Median for ID group=10150591 is 50, then if userid=30240768 has value by countlike on 25% more than the median of this group then "red".
=50/100*25=12.5 25% percentage from 50=12.5. So 50+12.5=62.5, If userid=30240768 has value more then 62.5 by countlike then "red"
i.e. userid=30240768 has value 99. so he is "red".
If userid has value on 25% less than the median by this group then "green". 50-12.5=37.5, here not such value. And last, if value in range ±24% from median for group then "orange". 24% from 50 =50/100*24=12, so if userid has value by countlike 50 ± 12 (38-62) then "orange".
So desired output
ID group UserID countlike median countlike
10150591 181078814 44 orange
10150591 88578209 50 orange
10150591 30240768 99 red
68837296 334686951 1 green
68837296 297170412 1 green
68837296 265332359 15 red
How do I comply with such conditions?

Answer 1

Here is an answer using dplyr. We aggregate the data to medians, merge the medians with the original data, and then calculate color.
First, we read the dput() data from the OP and remove rows that are missing.
data <- structure(list(ID.group = c(NA, 10150591L, NA, 10150591L, NA,
10150591L, NA, 68837296L, NA, 68837296L, NA, 68837296L, NA, 124771228L,
NA, 124771228L), UserID = c(NA, 181078814L, NA, 88578209L, NA,
30240768L, NA, 334686951L, NA, 297170412L, NA, 265332359L, NA,
216632504L, NA, 5272133L), countlike = c(NA, 44L, NA, 50L, NA,
99L, NA, 1L, NA, 1L, NA, 15L, NA, 41L, NA, 20L), statistics.snt = structure(c(1L,
2L, 1L, 2L, 1L, 2L, 1L, 2L, 1L, 2L, 1L, 2L, 1L, 2L, 1L, 2L), .Label = c("",
"fb"), class = "factor"), statistics.created_at = structure(c(1L,
8L, 1L, 4L, 1L, 7L, 1L, 2L, 1L, 2L, 1L, 5L, 1L, 3L, 1L, 6L), .Label = c("",
"10.04.2020 9:14", "11.04.2020 0:01", "11.04.2020 19:22", "12.04.2020 19:45",
"12.04.2020 6:54", "13.04.2020 20:47", "17.04.2020 23:02"), class = "factor"),
statistics.updated_at = structure(c(1L, 8L, 1L, 7L, 1L, 6L,
1L, 3L, 1L, 3L, 1L, 4L, 1L, 5L, 1L, 2L), .Label = c("", "22.04.2020 12:27",
"22.04.2020 12:51", "22.04.2020 14:19", "22.04.2020 5:41",
"22.04.2020 6:18", "22.04.2020 7:37", "30.04.2020 16:55"), class = "factor"),
statistics.is_recount = structure(c(1L, 2L, 1L, 2L, 1L, 2L,
1L, 2L, 1L, 2L, 1L, 2L, 1L, 2L, 1L, 2L), .Label = c("", "False"
), class = "factor")), class = "data.frame", row.names = c(NA,
-16L))
data <- data[!is.na(data$ID.group),]
Next, we load dplyr and calculate the desired output.
library(dplyr)
data %>% group_by(ID.group) %>%
summarise(.,mdn_countlike = median(countlike)) %>%
inner_join(.,data) %>%
mutate(color = case_when(countlike > 1.25 * mdn_countlike ~ "red",
countlike < 0.75 * mdn_countlike ~ "green",
countlike >= 0.75 * mdn_countlike &
countlike <= 1.25 * mdn_countlike ~ "orange")) -> mergedData
mergedData[,c("ID.group","UserID","countlike","mdn_countlike","color")]
...and the output:
> mergedData[,c("ID.group","UserID","countlike","mdn_countlike","color")]
# A tibble: 8 x 5
ID.group UserID countlike mdn_countlike color
<int> <int> <int> <dbl> <chr>
1 10150591 181078814 44 50 orange
2 10150591 88578209 50 50 orange
3 10150591 30240768 99 50 red
4 68837296 334686951 1 1 orange
5 68837296 297170412 1 1 orange
6 68837296 265332359 15 1 red
7 124771228 216632504 41 30.5 red
8 124771228 5272133 20 30.5 green
>