28.5. THE CENTRAL LIMIT THEOREM 777

Lemma 28.5.1 If {Xn}is a sequence of random vectors with values in Rpsuch that

limn→∞

φXn(t)≡ lim

n→∞φ λXn

(t) = ψ (t)

for all t, where ψ (0) = 1 and ψ is continuous at 0, then {λXn}∞

n=1 is tight.

In proving the central limit theorem, one considers the pointwise convergence of char-acteristic functions and then seeks to obtain information about the distribution of the limitfunction. In fact, one is in the situation of the following lemma which is Lemma 28.4.4.

Lemma 28.5.2 If φXn(t)→ φX (t) for all t, then whenever ψ ∈S,

λXn (ψ)≡∫Rp

ψ (y)dλXn (y)→∫Rp

ψ (y)dλX (y)≡ λX (ψ)

as n→ ∞.

The above gives what I want for ψ ∈ S but this needs to be generalized to ψ anybounded uniformly continuous function. The following is Lemma 28.4.5.

Lemma 28.5.3 If φXn(t)→ φX (t) , then if ψ is any bounded uniformly continuous

function,

limn→∞

∫Rp

ψdλXn =∫Rp

ψdλX .

Definition 28.5.4 Let µ be a Radon measure onRp. A Borel set A, is a µ continuityset if µ (∂A) = 0 where ∂A≡ A\ int(A) and int denotes the interior.

The main result is the following continuity theorem. More can be said about the equiv-alence of various criteria [6].

Theorem 28.5.5 If φXn(t)→ φX (t) then λXn (A)→ λX (A) whenever A is a λX

continuity set.

Proof: First suppose K is a closed set and let

ψk (x)≡ (1− k dist(x,K))+.

Thus, since K is closed limk→∞ψk (x) = XK (x). Choose k large enough that∫Rp

ψkdλX ≤ λX (K)+ ε.

Then by Lemma 28.5.3, applied to the bounded uniformly continuous function ψk,

lim supn→∞

λXn (K)≤ lim supn→∞

∫ψkdλXn =

∫ψkdλX ≤ λX (K)+ ε.

Since ε is arbitrary, this shows limsupn→∞ λXn (K)≤ λX (K) for all K closed.Next suppose V is open and let

ψk (x) = 1−(1− k dist

(x,VC))+.

28.5. THE CENTRAL LIMIT THEOREM 777Lemma 28.5.1 /f {X,,}is a sequence of random vectors with values in R? such thatlim $x, (t) = lim o,.., (t) = w(t)n—s0ofor all t, where y (0) = 1 and y is continuous at 0, then {A .x,,},_, is tight.In proving the central limit theorem, one considers the pointwise convergence of char-acteristic functions and then seeks to obtain information about the distribution of the limitfunction. In fact, one is in the situation of the following lemma which is Lemma 28.4.4.Lemma 28.5.2 If ¢ x, (t) > ¢x (t) for all t, then whenever w € 6,Ax, (v)= | v(y)dax,(») > [| wly)drx(v) =Ax (Wy)asn—- oo,The above gives what I want for y € G but this needs to be generalized to y anybounded uniformly continuous function. The following is Lemma 28.4.5.Lemma 28.5.3 If ¢x, (t) > x (€), then if y is any bounded uniformly continuousfunction,R? RPnooDefinition 28.5.4 Le: Lt be a Radon measure on R?. A Borel set A, is a UM continuityset if U (0A) = 0 where 0A = A \ int (A) and int denotes the interior.The main result is the following continuity theorem. More can be said about the equiv-alence of various criteria [6].Theorem 28.5.5 if 6x, (t) > ¢x (t) then dx, (A) + Ax (A) whenever Aisahxcontinuity set.Proof: First suppose K is a closed set and lety, (x) = (1—kdist(a,K))*.Thus, since K is closed limy_,..W; (@) = 2x (ax). Choose k large enough that/ wddx <Ax(K)+e.RPThen by Lemma 28.5.3, applied to the bounded uniformly continuous function y;,lim sup Ax, (K) <limsup | w,dax, = J marx <Ax (K)+€.n—yoo n—yooSince € is arbitrary, this shows limsup,,_,.. 2.x, (K) < Ax (K) for all K closed.Next suppose V is open and letW; (a) =1—(1—kdist (w,V°))”.