\(R_{xx}(m) = \frac{1}{N-|m|}\sum x(n)x(n+m)\) (unbiased) and the
divisor-N biased form. The unbiased estimate is NOT guaranteed
positive semi-definite, so feeding it to an AR solve can yield an
unstable model – which is why morie_yule_walker uses
the biased one. Mirrors morie.fn.rgacf.
Usage
morie_acf_estimate(x, max_lag = NULL)
Arguments
- x
numeric signal.
- max_lag
maximum lag; N - 1 by default.
Value
list: lags, acf_unbiased, acf_biased, N.
References
Rangayyan (2015), Ch. 3.
Examples
set.seed(1)
morie_acf_estimate(rnorm(50), max_lag = 5)$acf_unbiased
#> [1] 0.68748141 0.03104690 -0.10900106 -0.06796563 0.06613712 -0.15976493
