RunningCov: Fast Running Covariance Computation
In martakarass/convo: Fast Computation of Running Sample Statistics

Description Usage Arguments Details Value Examples

Computes running covariance between two sequences in a fixed width window, whose length corresponds to the length of the shorter sequence. Uses convolution implementation via Fast Fourier Transform.

1	RunningCov(x, y, circular = FALSE)

`x`	A numeric vector.
`y`	A numeric vector, of equal or shorter length than `x` sequence.
`circular`	Logical; whether running variance is computed assuming circular nature of `x` sequence (see Details).

Computes running covariance between two sequences in a fixed width window. The length of a window is equal to the shorter of the two sequences (y), and window "runs" over the length of longer sequence (x).

The length of output vector equals the length of x vector. Parameter circular determines whether x sequence is assumed to have a circular nature. Assume l_x is the length of sequence x, l_y is the length of shorter sequence y.

If circular equals TRUE then

first element of the output sequence corresponds to sample covariance between x[1:l_y] and y,
last element of the output sequence corresponds to sample covariance between c(x[l_x], x[1:(l_y - 1)]) and y.

If circular equals FALSE then

first element of the output sequence corresponds to sample covariance between x[1:l_y] and y,
the l_x - W + 1-th last element of the output sequence corresponds to sample covariance between x[(l_x - l_y + 1):l_x],
last W-1 elements of the output sequence are filled with NA.

A numeric vector.

x <- sin(seq(0, 1, length.out = 1000) * 2 * pi * 6)
y <- x[1:100]
out1 <- RunningCov(x, y, circular = TRUE)
out2 <- RunningCov(x, y, circular = FALSE)
plot(out1, type = "l"); points(out2, col = "red")