352 Consistency and Limiting Distributions
Σwhich is positive definite. Assume that the common moment generating function
M(t)exists in an open neighborhood of 0 .Let
Yn=
1
√
n
∑n
i=1
(Xi−μ)=
√
n(X−μ).
ThenYnconverges in distribution to aNp( 0 ,Σ)distribution.
Proof:Lett∈Rpbe a vector in the stipulated neighborhood of 0. The moment
generating function ofYnis
Mn(t)=E
[
exp
{
t′
1
√
n
∑n
i=1
(Xi−μ)
}]
= E
[
exp
{
1
√
n
∑n
i=1
t′(Xi−μ)
}]
= E
[
exp
{
1
√
n
∑n
i=1
Wi
}]
, (5.4.10)
whereWi=t′(Xi−μ). Note thatW 1 ,...,Wnare iid with mean 0 and variance
Var(Wi)=t′Σt. Hence, by the simple Central Limit Theorem,
1
√
n
∑n
i=1
Wi
D
→N(0,t′Σt). (5.4.11)
Expression (5.4.10), though, is the mgf of (1/
√
n)
∑n
i=1Wievaluated at 1. There-
fore, by (5.4.11), we must have
Mn(t)=E
[
exp
{
(1)
1
√
n
∑n
i=1
Wi
}]
→e^1
(^2) t′Σt/ 2
=et
′Σt/ 2
.
Because the last quantity is the moment generating function of aNp( 0 ,Σ) distri-
bution, we have the desired result.
SupposeX 1 ,X 2 ,...,Xnis a random sample from a distribution with mean
vectorμand variance-covariance matrixΣ.LetXnbe the vector of sample means.
Then, from the Central Limit Theorem, we say that
Xnhas an approximateNp
(
μ,^1 nΣ
)
distribution. (5.4.12)
A result that we use frequently concerns linear transformations. Its proof is
obtained by using moment generating functions and is left as an exercise.
Theorem 5.4.5.Let{Xn}be a sequence ofp-dimensional random vectors. Suppose
Xn
D
→ N(μ,Σ).LetAbe anm×pmatrix of constants and letb be anm-
dimensional vector of constants. ThenAXn+b
D
→N(Aμ+b,AΣA′).