Generating a order rank column with dplyr based on changes on the grouping variable

Question

I am having a little challenge with dplyr on generating a rank column on a tbl_df object from a log of transactions for a particular consumer. The data i have look like this:

                                        consumerid merchant_id      eventtimestamp merchant_visit_rank
                                              (chr)       (int)              (time)          (dbl)
            1  004a5cc3-3d60-4d14-85b3-706e454aae13          52 2015-01-15 13:33:00              0
            2  004a5cc3-3d60-4d14-85b3-706e454aae13          56 2015-01-16 13:58:03              1
            3  004a5cc3-3d60-4d14-85b3-706e454aae13          56 2015-01-16 13:58:41              0
            4  004a5cc3-3d60-4d14-85b3-706e454aae13          52 2015-01-16 13:59:05              1
            5  004a5cc3-3d60-4d14-85b3-706e454aae13          52 2015-01-16 13:59:55              1
            6  004a5cc3-3d60-4d14-85b3-706e454aae13          52 2015-01-16 14:15:56              0
            7  004a5cc3-3d60-4d14-85b3-706e454aae13          58 2015-01-21 13:52:18              1
            8  004a5cc3-3d60-4d14-85b3-706e454aae13          58 2015-01-21 13:52:19              0
            9  004a5cc3-3d60-4d14-85b3-706e454aae13          54 2015-01-21 13:52:24              0
            10 004a5cc3-3d60-4d14-85b3-706e454aae13          58 2015-01-21 13:52:29              0
            ..                                  ...         ...                 ...            ...

I want to generate a merchant visit rank so it tells me the order of this merchant during this transaction session. In our case the correct ranking would look :

                                        consumerid merchant_id      eventtimestamp merchant_visit_rank
                                              (chr)       (int)              (time)          (dbl)
            1  004a5cc3-3d60-4d14-85b3-706e454aae13          52 2015-01-15 13:33:00              1
            2  004a5cc3-3d60-4d14-85b3-706e454aae13          56 2015-01-16 13:58:03              2
            3  004a5cc3-3d60-4d14-85b3-706e454aae13          56 2015-01-16 13:58:41              2
            4  004a5cc3-3d60-4d14-85b3-706e454aae13          52 2015-01-16 13:59:05              3
            5  004a5cc3-3d60-4d14-85b3-706e454aae13          52 2015-01-16 13:59:55              3
            6  004a5cc3-3d60-4d14-85b3-706e454aae13          52 2015-01-16 14:15:56              3
            7  004a5cc3-3d60-4d14-85b3-706e454aae13          58 2015-01-21 13:52:18              4
            8  004a5cc3-3d60-4d14-85b3-706e454aae13          58 2015-01-21 13:52:19              4
            9  004a5cc3-3d60-4d14-85b3-706e454aae13          54 2015-01-21 13:52:24              5
            10 004a5cc3-3d60-4d14-85b3-706e454aae13          58 2015-01-21 13:52:29              6
            ..                                  ...         ...                 ...            ...

I have tried to play with the window functions in dplyr like this :

            measure_media_interaction %>% 
              #selecting the fields we wish from the dataframe
              select(consumerid,merchant_id,eventtimestamp) %>%
              #mutate a placeholder column to be used for the rank 
              mutate(merchant_visit = 0) %>% 
              #sort them by consumer and timestamp
              arrange(consumerid,eventtimestamp) %>%
              #change the column so it shows that this merchant was the first this consumer visited 
              #or not 
              mutate(merchant_visit = 
                       ifelse(lead(merchant_id)!=merchant_id,merchant_visit,merchant_visit+1))

However I am stuck and i don't know how to do it efficiently. Any ideas on this ?

Answer 1

Here is a solution. We use lag to test whether merchant_id changes and cumsum to increment the counter.

measure_media_interaction %>% 
  select(consumerid,merchant_id,eventtimestamp) %>%
  arrange(consumerid,eventtimestamp) %>%
  mutate(merchant_visit=cumsum(c(1,(merchant_id != lag(merchant_id))[-1])))