I can lag a vector like this, keeping/dublicating the initial value '1':
x <- c(1,0,0,0,0)
zoo::na.locf(lag(x,1), fromLast = T)
[1] 1 1 0 0 0
But if there is a leading '0' in the x vector, the following '1' is not repeated:
x <- c(0,1,0,0,0)
zoo::na.locf(lag(x,1), fromLast = T)
[1] 0 0 1 0 0
How can I repeat/keep the '1' if it is not leading? The result should look like this:
0 1 1 0 0
In other words: I want to fill the gap/lag between the original position of the first non-zero value and its lagged position with the first non-zero value. Here is a little more complex example:
c(0,0,0,1,2,3,0,0,0,0)
should become
c(0,0,0,1,1,2,3,0,0,0)
by lag 1
and
c(0,0,0,1,1,1,1,2,3,0)
by lag 3
Extract out the leading 0s, lag() the remaining vector with the first value as default, then add back the leading 0s. Wrapped in a helper function:
library(dplyr)
my_lag <- function(x, n = 1L) {
if (all(x == 0)) return(x)
x_0 <- x[cumall(x == 0)]
x_1 <- x[!cumall(x == 0)]
c(x_0, lag(x_1, n = n, default = x_1[[1]]))
}
Results with your example vectors:
x1 <- c(1,0,0,0,0)
x2 <- c(0,1,0,0,0)
x3 <- c(0,0,0,1,2,3,0,0,0,0)
my_lag(x1)
# 1 1 0 0 0
my_lag(x2)
# 0 1 1 0 0
my_lag(x3)
# 0 0 0 1 1 2 3 0 0 0
my_lag(x3, 3)
# 0 0 0 1 1 1 1 2 3 0