How to sum 5 minute intervals in R

Question

I have a dataset with precipitation records for every minute, for 6 different stations. I'd like to have summations for every 5 minutes, for every station. These are the first 5 rows of my dataset (in total I have 17280 rows):

  P_alex P_hvh P_merlijn P_pascal P_thurlede P_tosca                date
    0     0         0        0          0       0 2011-06-27 22:00:00
    0     1         5        2          0       0 2011-06-27 22:01:00
    0     0         0        0          0       0 2011-06-27 22:02:00
    0     6         2        3          0       0 2011-06-27 22:03:00
    0     0         0        0          0       0 2011-06-27 22:04:00

I tried to find help on the internet, but I can not find an answer that helps me.

I also needed houlry sums, for that I use the following code, but this code is useless if you want to make other summations

uur_alex = tapply(disdro$P_alex, as.POSIXct(trunc(disdro$date, "hour")), sum)

Now I would like a code I could use to make different summations, so for 5 minutes (as in the question), but also for half an hour. I hope somebody can help me.

Answer 1

you can use rollapply from the zoo package to achieve this. For example,

require(zoo)
tester <- data.frame(x=1:100,y=1:100)    
output <- rollapply(tester,5,(sum),by=5,by.column=TRUE,align='right')

Answer 2

cut works very nicely with date-time objects, and thus, can be used to create the 5 minute intervals you are hoping to aggregate over. Here's an example:

First, some sample data:

set.seed(1)
mydf <- data.frame(P_alex = sample(0:5, 40, replace = TRUE),
                   P_hvh = sample(0:3, 40, replace = TRUE),
                   date = as.POSIXct("2011-06-27 22:00:00") + 60 * 0:39)
list(head(mydf), tail(mydf))
# [[1]]
#   P_alex P_hvh                date
# 1      1     3 2011-06-27 22:00:00
# 2      2     2 2011-06-27 22:01:00
# 3      3     3 2011-06-27 22:02:00
# 4      5     2 2011-06-27 22:03:00
# 5      1     2 2011-06-27 22:04:00
# 6      5     3 2011-06-27 22:05:00
# 
# [[2]]
#    P_alex P_hvh                date
# 35      4     1 2011-06-27 22:34:00
# 36      4     3 2011-06-27 22:35:00
# 37      4     3 2011-06-27 22:36:00
# 38      0     1 2011-06-27 22:37:00
# 39      4     3 2011-06-27 22:38:00
# 40      2     3 2011-06-27 22:39:00

Now, perform your aggregation. In the following example, we aggregate all columns from the original dataset, but drop the "date" variable from the dataset (using mydf[setdiff(names(mydf), "date")]).

# Aggregate all columns by the intervals created with cut.
# For the dataset, we drop the original date column since
#   it is no longer needed here. Our function is "sum"
aggregate(. ~ cut(mydf$date, "5 min"), 
          mydf[setdiff(names(mydf), "date")], 
          sum)
#   cut(mydf$date, "5 min") P_alex P_hvh
# 1     2011-06-27 22:00:00     12    12
# 2     2011-06-27 22:05:00     16     8
# 3     2011-06-27 22:10:00     12     5
# 4     2011-06-27 22:15:00     17     6
# 5     2011-06-27 22:20:00     10     8
# 6     2011-06-27 22:25:00     11     8
# 7     2011-06-27 22:30:00     12     7
# 8     2011-06-27 22:35:00     14    13

Answer 3

One way is to map the dates to 5-minute blocks by using integer division (%/%). The base will be the UNIX epoch if using POSIXct datetimes. The you can sum on these blocks using aggregate.

x <- data.frame(date=Sys.time()+60*0:10,value1=0:10,value2=rnorm(11))

aggregate(.~as.numeric(date)%/%(5*60),data=x,FUN=sum)
  as.numeric(date)%/%(5 * 60)       date value1     value2
1                     4525797 1357739399      0  0.6209565
2                     4525798 6788697893     15 -1.4342917
3                     4525799 6788699393     40  0.8064627

Answer 4

If you are familiar with SQL, you can easily create SQL statement to group data into 5-minutes intervals. For example in postgresql you can use something like:

select Now(), date_trunc('hour',Now()) + interval '1 minute' * trunc(date_part('minute',Now())/5)*5

I use sqldf package to do all such transformations.