CODEWITHSUNDEEP
rsequencerle
Michel Mesquita
Michel Mesquita

Reputation: 783

How to combine rle with amount sum in R?

I have a dataset of 0s and 1s and amounts attached to them,

test = data.frame(seq = c(0,0,0,1,1,0,1,0,0), amount = c(91.0, 100.0, 0.0, 4.5, 5.5, 3.0, 23.0, 89.0, 56.0))



  seq amount
1   0   91.0
2   0  100.0
3   0    0.0
4   1    4.5
5   1    5.5
6   0    3.0
7   1   23.0
8   0   89.0
9   0   56.0

An event is defined by the first 0 in a subsequence of 0s. I am interested in the number of zeros (count) in each event, as well as the amount sums.

For test above, we would have:

So, I would like to create the following table,

|      Event |     count    |   amount |
|------------|--------------|----------|
|        1   |      3       | 191.0    |
|        2   |      1       | 3.0      |
|        3   |      2       | 145.0    |

In an earlier post, @27 ϕ 9 sent me this great suggestion for the Event and count columns.

with(rle(test), data.frame(id = sequence(sum(values == 0)), count = lengths[values == 0]))

But how can I add the amount sums still using rle?

Upvotes: 0

Views: 180

Answers (2)

GKi
GKi

Reputation: 39667

In case there is no urged need to use rle aggregate could be used like:

i <- which(test$seq == 0)
aggregate(cbind(count=1, amount=test$amount[i]),
          list(Event=cumsum(c(1, diff(i) > 1))), sum)
#  Event count amount
#1     1     3    191
#2     2     1      3
#3     3     2    145

Or rowsum:

i <- which(test$seq == 0)
rowsum(cbind(count=1, amount=test$amount[i]), cumsum(c(1, diff(i) > 1)))
#  count amount
#1     3    191
#2     1      3
#3     2    145

Upvotes: 0

Ronak Shah
Ronak Shah

Reputation: 388982

You may use data.table::rleid to create group of consecutive runs, count the number of rows in each group and sum the amount.

library(dplyr)

res <- test %>%
  group_by(Event = data.table::rleid(seq)) %>%
  summarise(seq = first(seq), 
            count = n(), 
            amount = sum(amount))

res

#  Event   seq count amount
#  <int> <dbl> <int>  <dbl>
#1     1     0     3    191
#2     2     1     2     10
#3     3     0     1      3
#4     4     1     1     23
#5     5     0     2    145

If you are interested only in 0 sequence -

res %>%
  filter(seq == 0) %>%
  mutate(Event = row_number()) %>%
  select(-seq)

#  Event count amount
#  <int> <int>  <dbl>
#1     1     3    191
#2     2     1      3
#3     3     2    145

If you are interested in continuing with the rle approach you can do -

with(rle(test$seq), data.frame(id = sequence(sum(values == 0)), 
                               count = lengths[values == 0], 
     amount = tapply(test$amount, rep(seq_along(values), lengths), sum)[values == 0]))

#  id count amount
#1  1     3    191
#3  2     1      3
#5  3     2    145

Upvotes: 1

Related Questions