7.6 Estimates ofthe Autocorrelation Function
L= prn(v(0) 18) pray(r(0) 8)... Pvex-n(vN — 1918) = FT pven (vti) 15)
“eae aed 60-9]
‘The logiLJis
log(L] = — ogame’) — BS 0 - 9)
A log{L]/as = 0 23 ow -s)
Thus, we have °
5a = a0 ix
ich is the sample mean estimate of the true mean, By previous results we know that this max-
‘sum likelihood estimate is unbiased and consistent
76 ESTIMATES OF THE AUTOCORRELATION FUNCTION
Two major estimators for the autocorrelation function are used on a regular basis. The
‘plications that call for autocorrelation estimates include data modeling, such as linear
pdiction, and power spectral estimation. We shall deal with both of these applications
Definition 7.15. The first estimate is called the unbiased autocorrelation esti
mate and is expressed as
5 1S
Rel) = Sy > XG 4 1kI) XG) forlk|