How to check for each country if date falls within specific interval across rows? - r

Question

Building on Check if a date is within an interval in R, we want to see if a specific event falls into a timeframe specified by another event. To give you a concrete example: For each country, did event (battle/protests/...) happen at the time of elections?
country <- c("Angola","Angola","Angola","Angola","Angola", "Benin","Benin","Benin","Benin","Benin","Benin")
event_type <- c("battle", "protests","riots", "riots", "elections","elections","protests","riots","violence","riots","elections")
event_date <- as.Date(c("2017-06-16", "2017-01-23", "2016-03-15", "2017-09-18", "2017-08-23", "2019-04-18", "2019-03-12", "2019-04-14", "2018-03-15", "2015-09-14", "2016-03-20"))
start_ecycle <- as.Date(c(NA,NA,NA,NA,"2017-05-25", "2019-01-18",NA,NA,NA,NA,"2015-12-21"))
end_ecycle <-as.Date(c(NA,NA,NA,NA,"2017-09-22","2019-05-18",NA,NA,NA,NA,"2016-04-19"))
mydata <- data.frame(country, event_type, event_date, start_ecycle, end_ecycle)
To this end, we created an interval variable
library(lubridate)
is.instant(mydata$start_ecycle); is.instant(mydata$end_ecycle)
mydata$ecycle <- interval(mydata$start_ecycle, mydata$end_ecycle)
Now, we got stuck. This is what the data.frame should look like in the end - i.e. here column G "ecycle_within" is added with 1 if event_date falls within ecycle (per country):
Any help much appreciated. Thanks!

Answer 1

Based on your comment about the elections cycles being across rows, I would recommend creating a separate dataset first with the elections data.
You can then join the election dates table. This will create a duplicate row for each event and election date range though.
The %within% lubridate function can then be used to check whether an event is within a specific election date range.
Lastly I reduce the number of rows, by filtering out rows corresponding to election date ranges that aren't relevant.
I am more familiar with dplyr and purrr and used them to implement it below. But you should be able to do something similar with base-r functions too.
I got the output close to your required output. But not 100% sure why you would like to do it this way.
library(tidyverse)
library(lubridate)
library(purrr)
elections <- mydata %>%
as_tibble() %>%
select(country, event_type, start_ecycle, end_ecycle) %>%
filter(event_type == "elections") %>%
mutate(election_year = year(start_ecycle)) %>%
select(country, start_ecycle, end_ecycle, election_year)
mydata2 <- mydata %>%
as_tibble() %>%
mutate(row = row_number()) %>%
select(row, country, event_type, event_date) %>%
left_join(elections, by = "country") %>%
mutate(ecycle = map2(start_ecycle, end_ecycle, ~ interval(.x, .y))) %>%
mutate(ecycle_within = map2_int(event_date, ecycle, ~ .x %within% .y)) %>%
select(-ecycle) %>%
group_by(country, event_type, event_date) %>%
arrange(desc(ecycle_within)) %>%
slice(1:1) %>%
ungroup() %>%
arrange(row) %>%
select(-row)
mydata2 %>% select(-election_year)
#> # A tibble: 11 x 6
#> country event_type event_date start_ecycle end_ecycle ecycle_within
#> <fct> <fct> <date> <date> <date> <int>
#> 1 Angola battle 2017-06-16 2017-05-25 2017-09-22 1
#> 2 Angola protests 2017-01-23 2017-05-25 2017-09-22 0
#> 3 Angola riots 2016-03-15 2017-05-25 2017-09-22 0
#> 4 Angola riots 2017-09-18 2017-05-25 2017-09-22 1
#> 5 Angola elections 2017-08-23 2017-05-25 2017-09-22 1
#> 6 Benin elections 2019-04-18 2019-01-18 2019-05-18 1
#> 7 Benin protests 2019-03-12 2019-01-18 2019-05-18 1
#> 8 Benin riots 2019-04-14 2019-01-18 2019-05-18 1
#> 9 Benin violence 2018-03-15 2019-01-18 2019-05-18 0
#> 10 Benin riots 2015-09-14 2019-01-18 2019-05-18 0
#> 11 Benin elections 2016-03-20 2015-12-21 2016-04-19 1