28.1. THE MULTIVARIATE NORMAL DISTRIBUTION 761

Proof: Consider E(eit·X) forX ∼ Np (m,Σ).

E(eit·X)≡ 1

(2π)p/2 (detΣ)1/2

∫Rp

eit·xe−12 (x−m)∗Σ−1(x−m)dx.

Let R be an orthogonal transformation such that

RΣR∗ = D = diag(σ

21, · · · ,σ2

p).

Let R(x−m) = y. Then

E(eit·X)= 1

(2π)p/2∏

pi=1 σ i

∫Rp

eit·(R∗y+m)e−12y∗D−1ydy.

ThereforeE(eit·X)= 1

(2π)p/2∏

pi=1 σ i

∫Rp

eis·(y+Rm)e−12y∗D−1ydy

where s= Rt. This equals

eit·mp

∏i=1

(∫R

eisiyie− 1

2σ2i

y2idyi

)1√

2πσ i

= eit·mp

∏i=1

(∫R

eisiσ iue−12 u2

du)

1√2π

= eit·mp

∏i=1

e−12 s2

i σ2i

1√2π

∫R

e−12 (u−isiσ i)

2du

By Lemma 28.0.1, this equals eit·me−12 ∑

pi=1 s2

i σ2i = eit·me−

12 t∗Σt. This proves 28.2.

SinceX1 andX2 are independent, eit·X1 and eit·X2 are also independent. Hence

E(eit·X1+X2

)= E

(eit·X1

)E(eit·X2

).

Thus,

E(eit·X1+X2

)= E

(eit·X1

)E(eit·X2

)= eit·m1e−

12 t∗Σ1teit·m2e−

12 t∗Σ2t

= eit·(m1+m2)e−12 t∗(Σ1+Σ2)t,

which, as shown above is the characteristic function of a random vector distributed asNp (m1 +m2,Σ1 +Σ2). Now it follows that X1 +X2 ∼ Np (m1 +m2,Σ1 +Σ2) by The-orem 27.1.4. This proves 28.1.

The assertion about −X is also easy to see because

E(

eit·(−X))

= E(

ei(−t)·X)

=1

(2π)p/2 (detΣ)1/2

∫Rp

ei(−t)·xe−12 (x−m)∗Σ−1(x−m)dx

=1

(2π)p/2 (detΣ)1/2

∫Rp

eit·xe−12 (x+m)∗Σ−1(x+m)dx

28.1. THE MULTIVARIATE NORMAL DISTRIBUTION 761Proof: Consider E (e’**) for X ~ N,(m, ).. 1 . 1 xyit-X it-2 .—5(a2—-m)*x' (2—m)re) (2m)?/? (det) !/? he c* aeLet R be an orthogonal transformation such thatRER* = D = diag (oj,--- ,0;,).Let R(a—m) = y. ThenE(et*) =i oF - [ eit (Rytm) o—jy"D ly dy,20 T]j=, 01 7?’ThereforeE (ci*X) _ 1 eis (ytRm) o—3y"D ly dy(20)? TH) 1 JR?where s = Rt. This equals; P ; ~Ly? 1amt [ere as)j=t WR 210;. p . 12 1_ elt'm [ ebiPil o— 34 au)IT( R V2. P . 2= etm Jie | e732 (u-is:0;) du; V2 JRi=]. . 1 2-2 . lax :By Lemma 28.0.1, this equals e™e~ 22/1597 = eit™e—3#'Zt_ This proves 28.2.Since X and X are independent, e!“*! and e'#*2 are also independent. HenceE (el X1+%2) —E (eP*'NE (et *2).Thus,E (cit X1+X2) -~E (ct *1) E (cit X2) _ gitm) pst Lit it-my ,—Ft*ot_ pit-(mi+my) ,—5t* (Zi +22)t9which, as shown above is the characteristic function of a random vector distributed asNp (m1 + mz,X1 + Xz). Now it follows that X | + X2 ~ N,(m;+m2,2; + X2) by The-orem 27.1.4. This proves 28.1.The assertion about —X is also easy to see becauseE (e*-*) E (e*)1 .Ff =e(22)?/? (detd)!/? l, °1 | it-« —4(a@+m)*2"!(a#+m)= eve 2 dx(2m)?/? (detd)!/? Jirp