5.5Normal Random Variables 169
The moment generating function of a normal random variable with parametersμand
σ^2 is derived as follows:
φ(t)=E[etX]
=
1
√
2 πσ
∫∞
−∞
etxe−(x−μ)
(^2) /2σ 2
dx
1
√
2 π
eμt
∫∞
−∞
etσye−y
(^2) /2
dy by lettingy=
x−μ
σ
eμt
√
2 π
∫∞
−∞
exp
{
−
[
y^2 − 2 tσy
2
]}
dy
eμt
√
2 π
∫∞
−∞
exp
{
−
(y−tσ)^2
2
- t^2 σ^2
2
}
dy
=exp
{
μt+
σ^2 t^2
2
}
1
√
2 π
∫∞
−∞
e−(y−tσ)
(^2) /2
dy
=exp
{
μt+
σ^2 t^2
2
}
(5.5.1)
where the last equality follows since
1
√
2 π
e−(y−tσ)
(^2) /2
is the density of a normal random variable (having parameterstσand 1) and its integral
must thus equal 1.
Upon differentiating Equation 5.5.1, we obtain
φ′(t)=(μ+tσ^2 ) exp
{
μt+σ^2
t^2
2
}
φ′′(t)=σ^2 exp
{
μt+σ^2
t^2
2
}
+exp
{
μt+σ^2
t^2
2
}
(μ+tσ^2 )^2
Hence,
E[X]=φ′(0)=μ
E[X^2 ]=φ′′(0)=σ^2 +μ^2
and so
E[X]=μ
Var(X)=E[X^2 ]−(E[X])^2 =σ^2
Thusμandσ^2 represent respectively the mean and variance of the distribution.