CODEWITHSUNDEEP
rsequences
RCW
RCW

Reputation: 35

How to find and label start / end points of unique sequences in R

I have a sequence of 1s and 0s alongside a time vector. I'd like to find the start and end time points of all the sequences of 1s, and give each sequence a unique ID. Here is some example data and my attempt so far.

Create dummy data

# Create the sequence
x = c(0,0,0,0,1,1,1,1,0,0,0,1,1,1,1,1,1,0,0,0,1,1,1,1,0)

# Create the time vector
t = 10:34

This is my effort

#Get changepoints using diff()
diff_result <- diff(x)

# Use ifelse() to get start and end times (i.e. on and off)
on_t <- ifelse(diff_result == 1, t, NA)
off_t <- ifelse(diff_result == -1, t, NA)

# Combine into data frame and remove NAs, add 1 to on_t
results <- data.frame(on_t  = on_t[!is.na(on_t)] + 1, off_t = off_t[!is.na(off_t)])

# Create unique ID for each sequence
results$ID <- factor(1:nrow(results))

print(results)

  on_t off_t ID
1   14    17  1
2   21    26  2
3   30    33  3

I'm sure there's a better way...

Upvotes: 1

Views: 535

Answers (3)

lmo
lmo

Reputation: 38510

Here is one method for finding the starting and stopping positions of the above vector:

# get positions of the 1s
onePos <- which(x == 1)
# get the ending positions
stopPos <- onePos[c(which(diff(onePos) != 1), length(onePos))]
# get the starting positions
startPos <- onePos[c(1, which(diff(onePos) != 1) + 1)]

The values of the ts can be obtained through subsetting:

t[startPos]
[1] 14 21 30
t[stopPos]
[1] 17 26 33

Finally, to add an id:

df <- data.frame(id=seq_along(startPos), on_t=t[startPos], off_t=t[stopPos])

Upvotes: 0

akuiper
akuiper

Reputation: 214977

Put the two vectors in a data.table and then do typical group by, filter and mutate transformation is another option:

library(data.table)
dt = data.table(seq = x, time = t)
dt[, .(on_t = min(time), off_t = max(time), lab = unique(seq)), .(ID = rleid(seq))]
  # Use rleid to create a unique ID for each sequence as a group by variable, find the start 
  # and end point for each sequence as well as a label for each sequence;
  [lab == 1]
  # filter label so that the result only contains time for sequence of 1
  [, `:=`(lab = NULL, ID = seq_along(ID))][]
  # Remove label and recreate the ID

#    ID on_t off_t
# 1:  1   14    17
# 2:  2   21    26
# 3:  3   30    33

Following OP's logic, which might be a better way:

d = diff(c(0, x, 0))
# prepend and append a 0 at the beginning and ending of x to make sure this always work 
# if the sequence starts or ends with 1.
results = data.frame(on_t = t[d == 1], off_t = t[(d == -1)[-1]])
# pick up the time where 1 sequence starts as on time, and 0 starts as off time. Here d is 
# one element longer than t and x but since the last element for d == 1 will always be false, it won't affect the result.
results$ID = 1:nrow(results)
# create an ID

results
#   on_t off_t ID
# 1   14    17  1
# 2   21    26  2
# 3   30    33  3

Upvotes: 3

Roman Luštrik
Roman Luštrik

Reputation: 70633

You can also do it this way.

x = c(0,0,0,0,1,1,1,1,0,0,0,1,1,1,1,1,1,0,0,0,1,1,1,1,0)
# Create the time vector
t = 10:34
xy <- data.frame(x, t)

mr <- rle(xy$x)$lengths
xy$group <- rep(letters[1:length(mr)], times = mr)

onesies <- xy[xy$x == 1, ]
out <- by(onesies, INDICES = onesies$group, 
             FUN = function(x) {
               data.frame(on_t = x$t[1], off_t = x$t[nrow(x)], ID = unique(x$group))
               })

do.call("rbind", out)

  on_t off_t ID
b   14    17  b
d   21    26  d
f   30    33  f

Upvotes: 2

Related Questions