28.3. FINDING MOMENTS 765

However, another way to prove this is to use Proposition 27.4.1 on Page 744 and con-sider the characteristic function. Let E (X j) = m j and P = ∑

pj=1 t jX j. Then since X is

normally distributed and the covariance is a diagonal,

D≡

 σ21 0

. . .0 σ2

p

,

E(eiP) = E

(eit·X)= eit·me−

12 t∗Σt = exp

(p

∑j=1

it jm j−12

t2j σ

2j

)

=p

∏j=1

exp(

it jm j−12

t2j σ

2j

)(28.6)

Also,

E(eit jX j

)= E

(exp

(it jX j + ∑

k ̸= ji0Xk

))= exp

(it jm j−

12

t2j σ

2j

)With 28.6, this shows E

(eiP)= ∏

pj=1 E

(eit jX j

)which shows by Proposition 27.4.1 that the

random variables,{

X1, · · · ,Xp}

are independent. ■

28.3 Finding MomentsLet X be a random variable with characteristic function

φ X (t)≡ E (exp(itX))

Then this can be used to find moments of the random variable assuming they exist. Thekth moment is defined as E

(Xk). This can be done by using the dominated convergence

theorem to differentiate the characteristic function with respect to t and then plugging int = 0. For example, φ

′X (t) = E (iX exp(itX)) and now plugging in t = 0 you get iE (X) .

Doing another differentiation you obtain φ′′X (t) = E

(−X2 exp(itX)

)and plugging in t = 0

you get −E(X2)

and so forth.An important case is where X is normally distributed with mean 0 and variance σ2.

In this case, as shown above, the characteristic function is e−12 t2σ2

. Also all moments existwhen X is normally distributed. So what are these moments? Dt

(e−

12 t2σ2

)=−tσ2e−

12 t2σ2

and plugging in t = 0 you find the mean equals 0 as expected.

Dt

(−tσ2e−

12 t2σ2

)=−σ

2e−12 t2σ2

+ t2σ

4e−12 t2σ2

and plugging in t = 0 you find the second moment is σ2. Then do it again.

Dt

(−σ

2e−12 t2σ2

+ t2σ

4e−12 t2σ2

)= 3σ

4te−12 t2σ2 − t3

σ6e−

12 t2σ2

Then E(X3)= 0.

Dt

(3σ

4te−12 t2σ2 − t3

σ6e−

12 t2σ2

)= 3σ

4e−12 t2σ2 −6σ

6t2e−12 t2σ2

+ t4σ

8e−12 t2σ2

28.3. FINDING MOMENTS 765However, another way to prove this is to use Proposition 27.4.1 on Page 744 and con-sider the characteristic function. Let E (X;) = mj and P = vie t;X;. Then since X isnormally distributed and the covariance is a diagonal, ,or 0D=0 O%,E iP - £E it-X itm —5et _ p a 1, 2(ce) = E(e**)=e*™e = exp D itimj— 3170)j=P 1= [Texo (im; — 3t03) (28.6)j=lAlso,1E (ei) —E (os ( + y? oxi) = exp (im; st03)kAjWith 28.6, this shows E (e P) =[[,_,£ (e ij iXi) which shows by Proposition 27.4.1 that therandom variables, {X10 +X Oa are » independent, |_|28.3 Finding MomentsLet X be a random variable with characteristic functionPx (t) SE (exp (itX))Then this can be used to find moments of the random variable assuming they exist. Thek" moment is defined as E (X Kk) . This can be done by using the dominated convergencetheorem to differentiate the characteristic function with respect to ¢ and then plugging int = 0. For example, @ (t) = E (iX exp (itX)) and now plugging in t = 0 you get iE (X).Doing another differentiation you obtain @{ (t) = E (—X? exp (itX)) and plugging int = 0you get —E (X ) and so forth.An important case is where X is normally distributed with mean 0 and variance 07.. we 12g? .In this case, as shown above, the characteristic function is e~ 2’ ° . Also all moments exist: . . 1,22 2 2when X is normally distributed. So what are these moments? D, (e-" ° )= tore 39and plugging in t = 0 you find the mean equals 0 as expected.D; (- pore te) — pene +Pote —$ro?and plugging in t = 0 you find the second moment is 07. Then do it again.1421,22 1,22 1,22 2D; (-o7e 4 Pote 29 ) =3o0'te 2" ° —Po%e 2[oyThen E (X*) =0.1,22 _ 2 142-2 _ 2D; (30%te 5P0? _ 365% 0") — 30% SPO? _ 6g 6 20-3? +to8e 1Po