Counting time between two events associated with a value column - r

Question

Description
I am trying to count time (in mins) between two events associated with a value column. The data has date (o3.date), time(o4.arrival), and a value (i1.arrivalEval) column, similar to the sample dataset below. The value column (i1.arrivalEval) represents a waiting time, and there could be two events associated with it.
First, if this value exceeds the 90th percentile of all waiting time, overcrowding happens. The overcrowding value is defined as the 90th percentile.
Second, once the overcrowding happens, it takes time to normalized to the average waiting time (average object in the below code) during the day. I am trying to count this time to normalized to the average waiting time after overcrowding for each incident and for each day.
For example, for day 2013-01-01, the waiting time (i1.arrivalEval) spiked to 179 mins (overcrowding is 176) at 15:30:00. Then, it reaches to below average (around 101) waiting time at 16:00:00. Thus, the count of minutes will be 30 mins. The second incident of overcrowding on the same day will be 2013-01-01 22:20:00 and it takes 10 mins to reach below the average waiting time. When one day has multiple overcrowding incidents, record each incident with the time of overcrowding.
If the computation can be done under ddply or other tidyverse packages, it would be great.
Desired Output
The output will have four columns: date, datetime of overcrowding, datetime of normalization (i.e., reaching to below average waiting time), and the time between two events.
Sample Dataset
# sample dataset
library(tidyverse)
library(tibble)
library(lubridate)
rename <- dplyr::rename
select <- dplyr::select
set.seed(10000)
id <- 1:1000
set.seed(10000)
i1.arrivalEval <- sample(10:200, 1000, replace = T)
set.seed(10003)
o4.arrival <- sample(seq(as.POSIXct('2013/01/01'), as.POSIXct('2013/01/30'), by="10 mins"), 1000)
d <-
data.frame(id, i1.arrivalEval, o4.arrival) %>%
mutate(o3.date = as.Date(o4.arrival)) %>%
arrange(o4.arrival)
average <- mean(d$i1.arrivalEval)
overcrowding <- quantile(d$i1.arrivalEval,.90)
# Goal: For *each day*, *counting number of minutes* until the waiting time reaches
# below average after overcrowding (spike in waiting time)
#
# For example, for day '2013-01-01', the waiting time (i1.arrivalEval) spiked to 179 mins (> overcrowding 176) at 15:30:00
# Then, it reaches to below average (around 101) waiting time at 16:00:00.
# Thus, the count of minutes will be 30 mins.
# The second incident of overcrowding on the day will be 2013-01-01 22:20:00 and
# it takes 10 mins to reach bewlow the average waiting time.
# When one day has multiple overcrowding incidents, record each incident with the time of overcrowding.
# The output will have four columns:
# date, time of overcrowding, time of resolved time, and count of minutes until below average waiting time
Update: updated dataset and the solution
# Updated dataset by #Donalid Seinen
set.seed(10000)
d <- data.frame(id = 1:1000, i1.arrivalEval = sample(10:200, 1000, replace = T),
o4.arrival = sample(seq(as.POSIXct('2013/01/01'),
as.POSIXct('2013/01/30'),
by="10 mins"), 1000)) %>%
mutate(o3.date = as.Date(o4.arrival)) %>%
arrange(o4.arrival)
average <- mean(d$i1.arrivalEval)
overcrowding <- quantile(d$i1.arrivalEval,.90)
# Solution
d %>%
ddply('o3.date', function(i){
i %>%
select(-o3.date) %>%
mutate(
o2.overcrowding = ifelse(i1.arrivalEval >= overcrowding, 1, 0),
o1.incidents = cumsum(o2.overcrowding)
) %>%
filter(o1.incidents > 0) %>%
mutate(
o1.slow = ifelse(i1.arrivalEval <= average, 1, 0)
) %>%
filter(o2.overcrowding != 0 | o1.slow != 0) %>%
arrange(o4.arrival) %>%
mutate(o4.normalized = lead(o4.arrival, 1L)) %>%
select(id, i1.arrivalEval, o4.arrival, o4.normalized, everything()) %>%
filter(o2.overcrowding == 1) %>%
mutate(o1.count = difftime(o4.normalized, o4.arrival, unit = 'mins')) %>%
select(-o2.overcrowding, -o1.incidents, -i1.arrivalEval, -o1.slow)
})