Our goal in this lecture is to develop asymptotic distribution
theory for the sample autocorrelation function.
We let and
be the ACF and estimated ACF
respectively.
We begin by reducing the behaviour of
to the behaviour
of
, the sample autocovariance. Our approach is standard
Talyor expansion.
Suppose you have pairs of random variables with
We want to compute the mean of this expression term by term and the
variance by using the formula for the variance of the sum and so on.
However, what we really do is truncate the infinite sum at some finite
number of terms and compute moments of the finite sum.
I want to be clear about the distinction; to do so I give an example.
Imagine that has a bivariate normal distribution with means
, variances
,
and
correlation
between
and
. The quantity
does not have a well defined mean because
E
. Our expansion is
still valid, however. Stopping the sum at
leads to the approximation
![]() |
![]() |
|
![]() |
Definition: : If is a sequence of random variables and
a sequence of
constants then we write
The idea is that
means that
is
proportional in size to
with the ``constant of
proportionality'' being a random variable which is not
likely to be too large. We also often have use for notation
indicating that
is actually small compared to
.
Definition: : We say
if
in probability:
for each
You can manipulate and
notation algebraically with a few rules:
These notions extend Landau's and
notation to
random quantities.
Example: : In our ratio example we have
![]() |
![]() |
|
![]() |
![]() |
![]() |
|
![]() |
||
![]() |
Keeping only terms of order
we find
To keep terms up to order
we have to keep terms out to
(In general
In order to compute the approximate variance we ought to compute the second moment
of
and subtract the square of the first moment.
Imagine you had a random variable of the form
Now I want to apply these ideas to estimation of . We make
be
and
be
(and replace
by
). Our first order approximation to
is
I now evaluate means and variances in the special case where has
been calculated using a known mean of 0. That is
To compute the variance we begin with the second moment which is
For and
or
the expectation is simply
while for
we get
.
Thus the variance of the sample variance (when the mean is known
to be 0) is
Having computed the variance it is usual to look at the large
sample distribution theory. For the usual central limit theorem
applies to
(in the case of white noise) to prove that
For the theorem needed is called the
-dependent central
limit theorem; it shows that
The sample autocorrelation at lag is
It is possible to verify that subtraction of from the
observations before computing the sample covariances does not
change the large sample approximations, although it does affect
the exact formulas for moments.
When the series is actually not white noise the situation is
more complicated. Consider as an example the model
There are versions of the central limit theorem called
mixing central limit theorems which can be used for ARMA() processes
in order to conclude that
The general tactic is the method or Taylor expansion. In this
case for each sample size
you have two estimates, say
and
of two parameters. You want distribution theory for the ratio
. The idea is to write
where
and then make use of the fact that
and
are
close to the parameters they are estimates of. In our case
is the sample autocovariance at lag
which is close to the
true autocovariance
while the denominator
is the
sample autocovariance at lag 0, a consistent estimator of
.
Write
![]() |
![]() |
![]() |
Similar ideas can be used for the estimated sample partial ACF.
In order to test the hypothesis that a series is white noise using the
distribution theory just given, you have to produce a single statistic
to base youre test on. Rather than pick a single value of the
suggestion has been made to consider a sum of squares or a weighted
sum of squares of the
.
A typical statistic is
When the parameters in an ARMA() have been estimated by maximum likelihood
the degrees of freedom must be adjusted to
. The resulting
test is the Box-Pierce test; a refined version which takes better account
of finite sample properties is the Box-Pierce-Ljung test. S-Plus plots the
-values from these tests for 1 through 10 degrees of freedom as
part of the output of arima.diag.