Code sequence by group in R: recurring values within group

Question

I would like to code a numerical sequence based on the combination of three columns: ID, year, and location. I want to number the years a person spent in one place as a sequence. The sequence should start anew in the year of a location change, so even if a person returns to a place that they have been before, the sequence should start anew.

The df with sequence should look like:

ID yr loc seq
1 1990 A 1
1 1991 A 2
1 1992 B 1
1 1993 B 2
1 1994 B 3
2 1990 B 1
2 1991 B 2
2 1992 A 1
2 1993 B 1
2 1994 B 2
3 1990 C 1
3 1991 C 2
3 1992 C 3
3 1993 B 1
3 1994 C 1

Do you have an idea on how to code this in R?

Ronak Shah · Accepted Answer · 2020-03-09T00:58:19.930

In data.table, we have rleid function which makes it simple here.

library(data.table)

setDT(df)[, seq1 := seq_len(.N), .(ID, rleid(loc))]

df
#    ID   yr loc seq seq1
# 1:  1 1990   A   1    1
# 2:  1 1991   A   2    2
# 3:  1 1992   B   1    1
# 4:  1 1993   B   2    2
# 5:  1 1994   B   3    3
# 6:  2 1990   B   1    1
# 7:  2 1991   B   2    2
# 8:  2 1992   A   1    1
# 9:  2 1993   B   1    1
#10:  2 1994   B   2    2
#11:  3 1990   C   1    1
#12:  3 1991   C   2    2
#13:  3 1992   C   3    3
#14:  3 1993   B   1    1
#15:  3 1994   C   1    1

We can use rleid in dplyr and base R approaches to get expected output.

library(dplyr)
df %>%
 group_by(ID, grp = data.table::rleid(loc)) %>%
  mutate(seq1 = row_number())

Or in base R :

df$seq1 <- with(df, ave(yr, ID, data.table::rleid(loc), FUN = seq_along))

A concise option suggested by @chinsoon12 is to use rowid function.

setDT(df)[, seq2 := rowid(ID, rleid(loc))]

data

df <- structure(list(ID = c(1L, 1L, 1L, 1L, 1L, 2L, 2L, 2L, 2L, 2L, 
3L, 3L, 3L, 3L, 3L), yr = c(1990L, 1991L, 1992L, 1993L, 1994L, 
1990L, 1991L, 1992L, 1993L, 1994L, 1990L, 1991L, 1992L, 1993L, 
1994L), loc = structure(c(1L, 1L, 2L, 2L, 2L, 2L, 2L, 1L, 2L, 
2L, 3L, 3L, 3L, 2L, 3L), .Label = c("A", "B", "C"), class = "factor"), 
seq = c(1L, 2L, 1L, 2L, 3L, 1L, 2L, 1L, 1L, 2L, 1L, 2L, 3L, 
1L, 1L)), class = "data.frame", row.names = c(NA, -15L))

and also in `data.table`, using `setDT(df)[, seq2 := rowid(ID, rleid(loc))]` — chinsoon12, Mar 09 '20 at 00:45

score 1 · Answer 2 · answered Mar 08 '20 at 17:41

In base R, we can use rle with ave

df$seq <- with(df, ave(yr, ID, with(rle(as.character(loc)), 
        rep(seq_along(values), lengths)), FUN = seq_along))
df$seq
#[1] 1 2 1 2 3 1 2 1 1 2 1 2 3 1

data

df <- structure(list(ID = c(1L, 1L, 1L, 1L, 1L, 2L, 2L, 2L, 2L, 2L, 
3L, 3L, 3L, 3L, 3L), yr = c(1990L, 1991L, 1992L, 1993L, 1994L, 
1990L, 1991L, 1992L, 1993L, 1994L, 1990L, 1991L, 1992L, 1993L, 
1994L), loc = structure(c(1L, 1L, 2L, 2L, 2L, 2L, 2L, 1L, 2L, 
2L, 3L, 3L, 3L, 2L, 3L), .Label = c("A", "B", "C"), class = "factor"), 
seq = c(1L, 2L, 1L, 2L, 3L, 1L, 2L, 1L, 1L, 2L, 1L, 2L, 3L, 
1L, 1L)), class = "data.frame", row.names = c(NA, -15L))

score 0 · Answer 3 · answered Mar 08 '20 at 11:20

A way with dplyr:

library(dplyr)

df %>% 
  group_by(ID, idx = cumsum(+(loc != lag(loc, default = first(loc))))) %>%
  mutate(seq = row_number()) %>%
  ungroup() %>% select(-idx)

Output:

# A tibble: 15 x 4
      ID    yr loc     seq
   <int> <int> <fct> <int>
 1     1  1990 A         1
 2     1  1991 A         2
 3     1  1992 B         1
 4     1  1993 B         2
 5     1  1994 B         3
 6     2  1990 B         1
 7     2  1991 B         2
 8     2  1992 A         1
 9     2  1993 B         1
10     2  1994 B         2
11     3  1990 C         1
12     3  1991 C         2
13     3  1992 C         3
14     3  1993 B         1
15     3  1994 C         1

Code sequence by group in R: recurring values within group

3 Answers3

data