R function for different follow-up intervals lubridate-CodePudding

I have the following df:

df = data.frame(id=c(1,1,1,1,1,1),
                date=c(as.Date("2000-01-01"), as.Date("2000-07-11"),
                       as.Date("2000-08-01"), as.Date("2000-12-31"),
                       as.Date("2002-05-04"), as.Date("2002-06-01")))

I need the following result:


result = data.frame(id=c(1,1,1,1,1,1),
                date=c(as.Date("2000-01-01"), as.Date("2000-07-11"),
                       as.Date("2000-08-01"), as.Date("2000-12-31"),
                       as.Date("2002-05-04"), as.Date("2002-06-01")),
                days_91 = c(0,0,1,0,0,1),
                days_182 = c(0,0,1,0,0,1),
                days_273 = c(0,1,1,1,0,1),
                days_365 = c(0,1,1,1,0,1))

Basically, for a certain date I want to know if a prior dates exists for the same ID in the last X days.

I supposed a lubridate fuction must exist but did not find it.

Result:

id	date	days_91	days_182	days_273	days_365
1	2000-01-01	0	0	0	0
1	2000-07-11	0	0	1	1
1	2000-08-01	1	1	1	1
1	2000-12-31	0	1	1	1
1	2002-05-04	0	0	0	0
1	2002-06-01	1	1	1	1

For instance, for row 3 there is a previous date in the last 91, 182, 273 and 365 days. However in row 2 there is no previous visit in the last 91 and 182 days

CodePudding user response：

Here is another option using map2 and map_dfc from purrr. After providing a given date and the previous date (in sorted order), you can compare the difference of these two values to all elements in a numeric vector (containing the number of days such as 91, 182, etc.).

library(tidyverse)

my_days <- c(91, 182, 273, 365)

df %>%
  group_by(id) %>%
  arrange(date, .by_group = T) %>%
  mutate(days = map2(
    date,
    lag(date, default = as.Date(-Inf)),
    \(x, y) {
      bind_cols(map_dfc(set_names(my_days, paste0("days_", my_days)), ~ (x - y < .x)))
    }
  )) %>%
  unnest(days)

Output

     id date       days_91 days_182 days_273 days_365
  <dbl> <date>       <int>    <int>    <int>    <int>
1     1 2000-01-01       0        0        0        0
2     1 2000-07-11       0        0        1        1
3     1 2000-08-01       1        1        1        1
4     1 2000-12-31       0        1        1        1
5     1 2002-05-04       0        0        0        0
6     1 2002-06-01       1        1        1        1

CodePudding user response：

We can use dplyr to iterate over a list of dates you want to check for, and will return 1 if any date in the 'date' column is present within the previous x days:

library(dplyr)

dates_check <- c(91, 192, 213, 365) # Dates we want to check

prev_dates <- function(prev_date){
  colname <- paste('days_', prev_date, sep='') # Dynamically create the column name
  df <<- df %>%
    group_by(id) %>% # Group our data by id
    rowwise() %>% # Perform rowwise operation
    mutate(!!colname := as.integer(any(df$date > date - prev_date & df$date < date))) 
}


lapply(dates_check, prev_dates)

# A tibble: 6 x 6
# Rowwise:  id
     id date       days_91 days_182 days_273 days_365
  <dbl> <date>       <int>    <int>    <int>    <int>
1     1 2000-01-01       0        0        0        0
2     1 2000-07-11       0        0        1        1
3     1 2000-08-01       1        1        1        1
4     1 2000-12-31       0        1        1        1
5     1 2002-05-04       0        0        0        0
6     1 2002-06-01       1        1        1        1