How to calculate time difference with previous row of a data.frame by group

Question

The problem I am trying to solve is that I have a data frame with a sorted POSIXct variable in it. Each row is categorized and I want to get the time differences between eac

Answer 1

Try this:

library(dplyr)
df %>%
  arrange(category, randtime) %>%
  group_by(category) %>%
  mutate(diff = randtime - lag(randtime),
         diff_secs = as.numeric(diff, units = 'secs'))

#   category            randtime             diff   diff_secs
#     (fctr)              (time)           (dfft)       (dbl)
# 1        A 2015-06-01 11:10:54         NA hours          NA
# 2        A 2015-06-01 15:35:04   4.402785 hours   15850.027
# 3        A 2015-06-01 17:01:22   1.438395 hours    5178.222
# 4        B 2015-06-01 08:14:46         NA hours          NA
# 5        B 2015-06-01 16:53:43 518.955379 hours 1868239.364
# 6        B 2015-06-01 17:37:48  44.090950 hours  158727.420

You may also want to add replace(is.na(.), 0) to the chain.

Answer 2

In base R you can use:

# creating an ordered data.frame
df <- data.frame(category, randtime)
df <- df[order(df$category, df$randtime),]

# calculating the timedifference
# option 1:
df$tdiff <- unlist(tapply(df$randtime, INDEX = df$category,
                          FUN = function(x) c(0, `units<-`(diff(x), "secs"))))
# option 2:
df$tdiff <- unlist(tapply(df$randtime, INDEX = df$category,
                          FUN = function(x) c(0, diff(as.numeric(x)))))

which gives:

> df
   category            randtime      tdiff
6         A 2015-06-01 11:10:54     0.0000
15        A 2015-06-01 15:35:04 15850.0271
18        A 2015-06-01 17:01:22  5178.2223
1         B 2015-06-01 08:14:46     0.0000
17        B 2015-06-01 16:53:43 31137.3227
19        B 2015-06-01 17:37:48  2645.4570
3         C 2015-06-01 10:09:50     0.0000
7         C 2015-06-01 12:46:40  9409.9693
9         C 2015-06-01 13:56:29  4188.4578
10        C 2015-06-01 14:24:18  1669.1326
12        C 2015-06-01 14:54:25  1807.1447
14        C 2015-06-01 15:05:07   641.7068
2         D 2015-06-01 09:28:16     0.0000
13        D 2015-06-01 14:55:40 19644.8313
4         E 2015-06-01 10:18:58     0.0000
5         E 2015-06-01 10:53:29  2071.2223
8         E 2015-06-01 13:26:26  9176.6263
11        E 2015-06-01 14:33:25  4019.0319
16        E 2015-06-01 15:57:16  5031.4183
20        E 2015-06-01 17:56:33  7156.8849

If you want minutes or hours, you can use "mins" or "hours" instead of "secs".

---

An alternative with the data.table package:

library(data.table)
# creating an ordered/keyed data.table
dt <- data.table(category, randtime, key = c("category", "randtime"))

# calculating the timedifference
# option 1:
dt[, tdiff := difftime(randtime, shift(randtime, fill=randtime[1L]), units="secs"), by=category]
# option 2:
dt[, tdiff := c(0, `units<-`(diff(randtime), "secs")), by = category]
# option 3:
dt[ , test := c(0, diff(as.numeric(randtime))), category]

which results in:

> dt
    category            randtime           tdiff
 1:        A 2015-06-01 11:10:54     0.0000 secs
 2:        A 2015-06-01 15:35:04 15850.0271 secs
 3:        A 2015-06-01 17:01:22  5178.2223 secs
 4:        B 2015-06-01 08:14:46     0.0000 secs
 5:        B 2015-06-01 16:53:43 31137.3227 secs
 6:        B 2015-06-01 17:37:48  2645.4570 secs
 7:        C 2015-06-01 10:09:50     0.0000 secs
 8:        C 2015-06-01 12:46:40  9409.9693 secs
 9:        C 2015-06-01 13:56:29  4188.4578 secs
10:        C 2015-06-01 14:24:18  1669.1326 secs
11:        C 2015-06-01 14:54:25  1807.1447 secs
12:        C 2015-06-01 15:05:07   641.7068 secs
13:        D 2015-06-01 09:28:16     0.0000 secs
14:        D 2015-06-01 14:55:40 19644.8313 secs
15:        E 2015-06-01 10:18:58     0.0000 secs
16:        E 2015-06-01 10:53:29  2071.2223 secs
17:        E 2015-06-01 13:26:26  9176.6263 secs
18:        E 2015-06-01 14:33:25  4019.0319 secs
19:        E 2015-06-01 15:57:16  5031.4183 secs
20:        E 2015-06-01 17:56:33  7156.8849 secs