7.2 Ergodic Random Processes
1 dy
= ime XO)
is
= tim? D x6
() > = Jims 2 (a)
Another convention is called the two-sided average, which is denoted as
=X()> = fim [x at
ile the respective convention for discrete random processes is either
a> ~fag7 2a
oo
> x@) waz
By
X00 = tins
Perhaps the two-sided convention is more common, especially for theoretical dei
ns, Consequently, we will use this convention in the next section. As we develops.
eos for parameters for discrete random sequences, we will adopt a notation similar
Se form in [7.1.4] without the limit operation, since we will be dealing with equa-
= for estimators that will depend on the use of Site length data sequences, This wl
me more appaeat ltr.
ERGODIC RANDOM PROCESSES
SS= previous chapters we have assumed that the statistical properties of a random
=s were given. However, in practice such knowledge is often not available. Con-
ly, we need to develop procedures for acquiring or estimating the statistical pa-
of interest by using measurements taken from a member function of a random
>. It turns out that for many stationary random processes, we can use time aver-
2 member function in place of ensemble averages ofthe random process. For ex-
under certain circumstances, we can use the sample mean, discussed above, in
the mean value calculated by ensemble averaging, When we can use time aver-
= place of ensemble averages, we have what is referred to loosely as an ergodie
Process. It tums out that there are a number of ergodic theorems that mathe-
y define ergodicity. We will discuss a few of these theorems,
* ergodic theorem comes from the strong law of large numbers and says that
= is « wide sense stationary, discrete random sequence with independent, iden-
Sstributed random variables with finite mean jtx, then the time average of the
sce of a member function will converge to the ensemble average with proba