grouping continuous data with specific pattern - r

Question

I have a data frame with a column like this (I am not posting other columns)
Value
1
1
1
0
0
1
0
0
1
1
2
2
0
0
1
0
0
1
1
1
0
0
2
2
1
1
2
0
0
1
0
I am trying to group it based on a specific condition. Grouping has to be done when I have 1 and 2. But conditions like these are one group :
1 1 0 0 1 1 0 0
Basically I need to group occurrences of 1 but in between 0s are allowed
Expected output:
Value Group
1 1
1 1
1 1
0 1
0 1
1 1
0 1
0 1
1 1
1 1
2 2
2 2
0 2
0 2
1 3
0 3
0 3
1 3
1 3
1 3
0 3
0 3
2 4
2 4
1 5
1 5
2 6
0 6
0 6
1 7
0 7
2 8
0 8
2 8
1 9

Answer 1

Here is another option using data.table:
DT[, Group := .GRP, .(date, rleid(nafill(replace(Value, Value==0L, NA_integer_), "locf")))]

Answer 2

Here is another base approach that uses ave() to count the changes between 1 and 2 and then uses cummax() on the result to give the final groupings.
dat$Group <- cummax(ave(dat$Value, dat$Value == 0, FUN = function(x) cumsum(c(x[1], diff(x) != 0))))
dat
Value Group
1 1 1
2 1 1
3 1 1
4 0 1
5 0 1
6 1 1
7 0 1
8 0 1
9 1 1
10 1 1
11 2 2
12 2 2
13 0 2
14 0 2
15 1 3
16 0 3
17 0 3
18 1 3
19 1 3
20 1 3
21 0 3
22 0 3
23 2 4
24 2 4
25 1 5
26 1 5
27 2 6
28 0 6
29 0 6
30 1 7
31 0 7
In response to your comment, if you want the result by grouped by date, you can use a nested ave():
ave(ave(dat$Value, dat$Value == 0, dat$date, FUN = function(x) cumsum(c(x[1], diff(x) != 0))), dat$date, FUN = cummax)

Answer 3

This loop in Base-R does the trick
group <- 0
lastgroupvalue <- NA
data$Group <- NA
for(i in 1:nrow(data)){
if(!data$Value[i] %in% c(lastgroupvalue, 0)){
group <- group + 1
lastgroupvalue <- data$Value[i]
}
data$Group[i] <- group
}
> data
Value Group
1 1 1
2 1 1
3 1 1
4 0 1
5 0 1
6 1 1
7 0 1
8 0 1
9 1 1
10 1 1
11 2 2
12 2 2
13 0 2
14 0 2
15 1 3
16 0 3
17 0 3
18 1 3
19 1 3
20 1 3
21 0 3
22 0 3
23 2 4
24 2 4
25 1 5
26 1 5
27 2 6
28 0 6
29 0 6
30 1 7
31 0 7
Data:
data <- structure(list(Value = c(1L, 1L, 1L, 0L, 0L, 1L, 0L, 0L, 1L,
1L, 2L, 2L, 0L, 0L, 1L, 0L, 0L, 1L, 1L, 1L, 0L, 0L, 2L, 2L, 1L,
1L, 2L, 0L, 0L, 1L, 0L)), class = "data.frame", row.names = c(NA,
-31L))

Answer 4

Another solution that avoids loops, that works similar to Limey's solution but uses cumsum to create the groups.
df$Group <- dplyr::na_if(df$Value, 0)
df <- tidyr::fill(df, Group, .direction = "down")
df$Group <- cumsum(df$Group != dplyr::lag(df$Group, default = -1))
> df
Value Group
1 1 1
2 1 1
3 1 1
4 0 1
5 0 1
6 1 1
7 0 1
8 0 1
9 1 1
10 1 1
11 2 2
12 2 2
13 0 2
14 0 2
15 1 3
16 0 3
17 0 3
18 1 3
19 1 3
20 1 3
21 0 3
22 0 3
23 2 4
24 2 4
25 1 5
26 1 5
27 2 6
28 0 6
29 0 6
30 1 7
31 0 7

Answer 5

Or a tidyverse solution that avoids loops:
x <- tibble(Value=c(1,1,1,0,0,1,0,0,1,1,2,2,0,0,1,0,0,1,1,1,
0,0,2,2,1,1,2,0,0,1,0,2,0,2,1)) %>%
mutate(ModValue=ifelse(Value == 0, NA, Value)) %>%
fill(ModValue, .direction="down")
runLengths <- rle(x$ModValue)
groupIndex <- unlist(lapply(1:length(runLengths$lengths),
function(x) rep(x, runLengths$lengths[x]))
)
x <- x %>% add_column(Group=groupIndex) %>% select(-ModValue)
Your input data has a different length to your expected output. Took me a while to work that out... :)
** Edit **
And an inelegant solution to account for changing days (or other super-groupings...
x <- tibble(
RowNumber=1:35,
Date=lubridate::ymd(c(rep("2020-05-31", 20), rep("2020-06-01", 15))),
Value=c(1,1,1,0,0,1,0,0,1,1,2,2,0,0,1,0,0,1,1,1,0,0,2,2,1,1,2,0,0,1,0,2,0,2,1))
# Check we have a change of date mid-sequence
x %>% filter(row_number() > 15 & row_number() < 25)
x <- x %>%
mutate(ModValue=ifelse(Value == 0, NA, Value)) %>%
fill(ModValue, .direction="down")
# Inelegantly compute the groups
make_groups <- function(x) {
runs <- rle(x)
return(tibble(GroupWithinDay=unlist(
lapply(1:length(runs$lengths),
function(x) rep(x, runs$lengths[x])))))
}
y <- x %>% group_by(Date) %>% do(make_groups(.$ModValue))
x <- x %>% add_column(GroupWithinDay=y$GroupWithinDay) %>% select(-ModValue)
# Check the change of date is handled correctly
x %>% filter(row_number() > 15 & row_number() < 25)
Giving
# A tibble: 9 x 4
RowNumber Date Value GroupWithinDay
<int> <date> <dbl> <int>
1 16 2020-05-31 0 3
2 17 2020-05-31 0 3
3 18 2020-05-31 1 3
4 19 2020-05-31 1 3
5 20 2020-05-31 1 3
6 21 2020-06-01 0 1
7 22 2020-06-01 0 1
8 23 2020-06-01 2 2
9 24 2020-06-01 2 2