CODEWITHSUNDEEP
r
otwtm
otwtm

Reputation: 1999

Check R data.frame column for equal values in another column

I'm looking for a vectorized solution to the following problem. There are customers that can have one of two different products, x or y, at a time. I would like to identify all rows of product x that are followed by product y for the same customer. In that case, the to_date of product x would be the same as the from_date of product y. Here is an example:

customerid = c(rep(1,2),rep(2,3))
product = c("x", "y", "x", "x", "y")
from_date = as.Date(c("2000-01-01", "2000-06-07","2001-02-01","2005-01-01","2005-11-01"))
to_date = as.Date(c("2000-06-07", "2000-10-31","2002-04-01","2005-11-01","2006-01-01"))

data.frame(customerid, product, from_date, to_date)

      customerid product  from_date    to_date
1          1       x 2000-01-01 2000-06-07
2          1       y 2000-06-07 2000-10-31
3          2       x 2001-02-01 2002-04-01
4          2       x 2005-01-01 2005-11-01
5          2       y 2005-11-01 2006-01-01

The desired output would look like:

  customerid product  from_date    to_date followed_by_y
1          1       x 2000-01-01 2000-06-07             yes
2          1       y 2000-06-07 2000-10-31             no
3          2       x 2001-02-01 2002-04-01             no
4          2       x 2005-01-01 2005-11-01             yes
5          2       y 2005-11-01 2006-01-01             no

My approach so far is to group the data.frame by costumerid with dplyr. But then I do not know how to check the to_date for equal values in the from_date.

Upvotes: 0

Views: 663

Answers (1)

arg0naut91
arg0naut91

Reputation: 14764

You could check for all conditions like below:

library(dplyr)

df %>%
  group_by(customerid) %>%
  mutate(followed_by_y = c('no', 'yes')[(product == 'x' &
                                         lead(product) == 'y' &
                                         to_date == lead(from_date)) + 1])

Output:

# A tibble: 5 x 5
# Groups:   customerid [2]
  customerid product from_date  to_date    followed_by_y
       <dbl> <fct>   <date>     <date>     <chr>        
1          1 x       2000-01-01 2000-06-07 yes          
2          1 y       2000-06-07 2000-10-31 no           
3          2 x       2001-02-01 2002-04-01 no           
4          2 x       2005-01-01 2005-11-01 yes          
5          2 y       2005-11-01 2006-01-01 no

Note, this is essentially the same as saying:

library(dplyr)

df %>%
  group_by(customerid) %>%
  mutate(followed_by_y = case_when(
    product == 'x' & lead(product) == 'y' & to_date == lead(from_date) ~ 'yes',
    TRUE ~ 'no')
  )

Upvotes: 1

Related Questions