Reputation: 2010
I have a series of data in the format (true/false). eg it looks like it can be generated from rbinom(n, 1, .1). I want a column that represents the # of rows since the last true. So the resulting data will look like
true/false gap
0 0
0 0
1 0
0 1
0 2
1 0
1 0
0 1
What is an efficient way to go from true/false to gap (in practice I'll this will be done on a large dataset with many different ids)
Upvotes: 1
Views: 76
Reputation: 132626
DF <- read.table(text="true/false gap
0 0
0 0
1 0
0 1
0 2
1 0
1 0
0 1", header=TRUE)
DF$gap2 <- sequence(rle(DF$true.false)$lengths) * #create a sequence for each run length
(1 - DF$true.false) * #multiply with 0 for all 1s
(cumsum(DF$true.false) != 0L) #multiply with zero for the leading zeros
# true.false gap gap2
#1 0 0 0
#2 0 0 0
#3 1 0 0
#4 0 1 1
#5 0 2 2
#6 1 0 0
#7 1 0 0
#8 0 1 1
The cumsum
part might not be the most efficient for large vectors. Something like
if (DF$true.false[1] == 0) DF$gap2[seq_len(rle(DF$true.false)$lengths[1])] <- 0
might be an alternative (and of course the rle
result could be stored temporarly to avoid calculating it twice).
Upvotes: 4
Reputation: 2275
Ok, let me put this in answer
1) No brainer method
data['gap'] = 0
for (i in 2:nrow(data)){
if data[i,'true/false'] == 0{
data[i,'gap'] = data[i-1,'gap'] + 1
}
}
2) No if check
data['gap'] = 0
for (i in 2:nrow(data)){
data[i,'gap'] = (data[i-1,'gap'] + 1) * (-(data[i,'gap'] - 1))
}
Really don't know which is faster, as both contain the same amount of reads from data, but (1) have an if statement, and I don't know how fast is it (compared to a single multiplication)
Upvotes: 0