772 CHAPTER 28. THE NORMAL DISTRIBUTION

Consequently,∣∣∣∣∫ φdµk−∫

φdλ

∣∣∣∣≤ ∣∣∣∣∫KCn

φdµk +∫

Kn

φdµk−(∫

Kn

φdλ +∫

KCn

φdλ

)∣∣∣∣≤∣∣∣∣∫Kn

φdµk−∫

Kn

φdλ n

∣∣∣∣+ ∣∣∣∣∫KCn

φdµk−∫

KCn

φdλ

∣∣∣∣≤

∣∣∣∣∫Kn

φdµk−∫

Kn

φdλ n

∣∣∣∣+ ∣∣∣∣∫KCn

φdµk

∣∣∣∣+ ∣∣∣∣∫KCn

φdλ

∣∣∣∣≤

∣∣∣∣∫Kn

φdµk−∫

Kn

φdλ n

∣∣∣∣+ Mn+

Mn

First let n be so large that 2M/n < ε/2 and then pick k large enough that the above expres-sion is less than ε.

Now suppose µn converges to λ weakly. Then for ε there is a compact set such thatλ (K) > 1− ε/2. This is true because of Lemma 9.8.5 on Page 255 which says that finitemeasures on a Polish space are inner regular. Then let ψ be a continuous function withvalues in [0,1] which equals 1 on K and is 0 off a compact set K̂⊇K. Then

∫ψdλ > 1−ε/2

and also, there exists N such that for all n ≥ N,∫

ψdµn > 1− ε/2. Thus n ≥ N impliesµn(K̂)> 1− ε/2. Therefore, enlarging K̂ finitely many times, one obtains K̃ ⊇ K such

that for all µn and λ ,λ(K̃),µn

(K̃)> 1− ε/2. Thus µn

(K̃C)≤ ε/2 < ε for all n and so

{µn} is tight as claimed. ■

Definition 28.4.9 Let µ,{µn} be probability measures defined on the Borel sets ofRp and let the sequence of probability measures, {µn} satisfy

limn→∞

∫φdµn =

∫φdµ.

for every φ a bounded continuous function. Then µn is said to converge weakly to µ .

With the above, it is possible to prove the following amazing theorem of Levy.

Theorem 28.4.10 Suppose {µn} is a sequence of probability measures defined on

the Borel sets of Rp and let{

φ µn

}denote the corresponding sequence of characteristic

functions. If there exists ψ which is continuous at 0, ψ (0) = 1, and for all t,

φ µn(t)→ ψ (t) ,

then there exists a probability measure λ defined on the Borel sets of Rp and

φ λ (t) = ψ (t) .

That is, ψ is a characteristic function of a probability measure. Also, {µn} convergesweakly to λ .

Proof: By Lemma 28.4.3 {µn} is tight. Therefore, there exists a subsequence{

µnk

}converging weakly to a probability measure λ which implies that

φ λ (t)≡∫

eit·xdλ (x) = limn→∞

∫eit·xdµnk

(x) = limn→∞

φ µnk(t) = ψ (t) ■

7712 CHAPTER 28. THE NORMAL DISTRIBUTIONConsequently,[edu [ oa2| < [ida [dai (joa + [, aa)<|[, edu J. dar +f, edn | oaa|< / odu,— | oddn|+| [ odus| +|[oaa|Kn Kn Ky KyM M< J bdyty— | dddq| += +=Kn Kn n nFirst let n be so large that 2M//n < €/2 and then pick k large enough that the above expres-sion is less than €.Now suppose [,, converges to A weakly. Then for € there is a compact set such thatA (K) > 1—€/2. This is true because of Lemma 9.8.5 on Page 255 which says that finitemeasures on a Polish space are inner regular. Then let y be a continuous function withvalues in [0, 1] which equals 1 on K and is 0 off a compact set K D K. Then f wda > 1—e/2and also, there exists N such that for alln > N,f wdu, > 1—€/2. Thus n > N impliesLL, (K ) >1-—e/2. Therefore, enlarging K finitely many times, one obtains K D K suchthat for all u,, and A, (K) ,u, (K) > 1—e/2. Thus pw, (K°) < €/2 <eé for all n and so{u,,} is tight as claimed.Definition 28.4.9 Le: Lt, {L,} be probability measures defined on the Borel sets ofR? and let the sequence of probability measures, {, } satisfylim [ edu, = [ ody.for every @ a bounded continuous function. Then LL, is said to converge weakly to [.With the above, it is possible to prove the following amazing theorem of Levy.Theorem 28.4.10 Suppose {[1,,} is a sequence of probability measures defined onthe Borel sets of R? and let {6 u, ¢ denote the corresponding sequence of characteristicnfunctions. If there exists w which is continuous at 0, y (0) = 1, and for all t,9, (t) > w(t),then there exists a probability measure 4 defined on the Borel sets of R? and$, (t) = w(t).That is, W is a characteristic function of a probability measure. Also, {U,,} convergesweakly to A.Proof: By Lemma 28.4.3 {,} is tight. Therefore, there exists a subsequence { Ln}converging weakly to a probability measure A which implies thatn-oo n-oo6, (t)= petra (x) = tim fe**au,, (x) = lim >, (t) = y(t) i