EE 562a
Homework Solutions 6
Wednesday 4 April 2007
1
1. Each of the convergence results deals with the following term
S
n
(
u
) =
n
i
=1
x
i
(
u
)
,
where
{
x
i
(
u
)
}
∞
i
=1
is a sequence of independent, identically distributed (iid) random variables,
each with mean
m
and variance
σ
2
. It follows that
m
n
=
E
{
S
n
(
u
)
}
=
nm
σ
2
n
=
V
ar
{
S
n
(
u
)
}
=
nσ
2
.
All three of these results are found in any good text on probability theory.
(a) The Weak Law of Large Numbers (WLLN) says the that the sample mean converges to
the mean in probability =
⇒
lim
n
→∞
P
1
n
S
n
(
u
)

m >
= 0
,
or
y
n
(
u
) =
1
n
S
n
(
u
)
→
m
=
y
(
u
)
in probability.
(b) The Strong Law of Large Numbers (SLLN) says the that the sample mean converges to
the mean almost surely =
⇒
P
lim
n
→∞
1
n
S
n
(
u
) =
m
= 1
,
or
y
n
(
u
) =
1
n
S
n
(
u
)
→
m
=
y
(
u
)
almost surely.
(c) The Central Limit Theorem is a statement about convergence in distribution. It says
that the cumulative effect of many iid disturbances is Gaussian.
Mathematically, we
have that the sequence of random variables
y
n
(
u
) =
S
n
(
u
)

nm
σ
√
n
converge in distribution to a mean zero unit variance Gaussian. In other words
lim
n
→∞
F
y
n
(
u
)
(
z
) =
z
∞
1
√
2
π
e

v
2
2
dv.
Another way of saying this is
y
n
(
u
) =
S
n
(
u
)

nm
σ
√
n
→
y
(
u
)
in distribution,
where
y
(
u
) is any Gaussian random variable zero mean and unit variance.
This preview has intentionally blurred sections. Sign up to view the full version.
View Full Document
2
EE 562a
Homework Solutions 6
Wednesday 4 April 2007
(d) We know that almost sure convergence always implies convergence in probability. Thus,
if we knew that the Strong Law of Large Numbers is true, we could conclude that the
Weak Law of Large Numbers is also true. If we only knew that the WLLN was true, we
could not conclude that the SLLN holds. In this sense, the SLLN is “stronger” than the
WLLN. You have probably all seen a proof of the WLLN, but probably none of you saw
the proof of the SLLN in you probability class.
In class we also proved another “law of large numbers,” namely
y
n
(
u
) =
1
n
S
n
(
u
)
→
m
=
y
(
u
)
in the mss.
It is easy to show that this law implies the WLLN (see problem 3). In general almost sure
convergence and convergence in the mss do not imply one another. So an appropriate
name for this law would be the “Equally Strong Law of Large Numbers.”
A final comment about these rules is that many of the hypotheses can be relaxed; for ex
ample we can prove the “Equally Strong Law of Large Numbers” and thus the WLLN with
the assumption that the
x
n
(
u
)’s are only uncorrelated (not independent).
Depending on
how advanced the probability text that you used was, you may have statement with weaker
hypotheses. For example, see Stark and Woods version of the Central Limit Theorem.
2. Solution:
Consider the space of all squareintegrable complex functions on [

1
/
2
,
1
/
2):
that is, all
functions
f
(
t
) such that
1
/
2

1
/
2

f
(
t
)

2
dt <
∞
. We can define an inner product (
f, g
) on this
space:
(
f, g
) =
1
/
2

1
/
2
f
*
(
t
)
g
(
t
)
dt.
This is the end of the preview.
Sign up
to
access the rest of the document.
 Spring '07
 ToddBrun
 Fourier Series, Trigraph

Click to edit the document details