Chapter 7 The Central Limit Theorem k k 7.1. Since C1 c (R ) ⇢ Cb (R ), it suffices to prove that if Z Z k lim f dµn = f dµ for all f 2 C1 c (R ), n!1 then µn ) µ. k Claim. For every f 2 Cc (Rk ) and ✏ 2 (0 , 1), there exists g✏ 2 C1 c (R ) such that supx |f(x) - g✏ (x)| 6 ✏. This Claim does the job. Indeed, the Claim implies that Z Z f dµn - g✏ dµn 6 ✏µn (Rk ) = ✏, and Z Z f dµ - g✏ dµ 6 ✏. Consequently, lim sup n!1 Z Z f dµn - f dµ 6 2✏. Since ✏ > 0 and f 2 Cc (Rk ) are arbitrary, the result follows. Now by Fejer’s theorem, if f 2 Cc (Rk ) then for all ✏ > 0 there exists h✏ 2 C1 (Rk ) such that supx |f(x) - h✏ (x)| 6 ✏/2. Suppose f(x) = 0 when k kxk > p. Then it suffices to prove that there exists g✏ 2 C1 c (R ) such that supkxk6p |g✏ (x) - h✏ (x)| 6 ✏. But this is easy. For instance, let ✓ ◆ 1 (x) := exp - 2 . (p - kxk)+ k Then, 2 C1 (x) 6 1 for all x, and (x) > exp(-1/p(p c (R ), 0 6 1)) > 1 - ✏ when kxk < p, if p = p(✏) is chosen sufficiently large. Then g✏ (x) := (x)h✏ (x) satisfies (1 - ✏)h✏ (x) 6 g✏ (x) 6 h✏ (x) 17 when kxk < p. 18 CHAPTER 7. THE CENTRAL LIMIT THEOREM Therefore, ◆ sup |g✏ (x) - h✏ (x)| 6 max sup |h✏ (x)| , 1 ✏ 6 K✏, kxk6p ✓ x where K := supx |f(x)| + ✏ 6 supx |f(x)| + 1. 7.2. Let X have distribution µ. Then the characteristic function of -X is ⇤ ⇥ b (t). E e-itX = E[cos(tX)] - itE[sin(tX)] = E [eitX ] = µ b is a characteristic function. Therefore, µ Let Y be independent from X, and have the distribution P{Y = 1} = P{Y = -1} = 1/2. Define Z = XY. Then, the characteristic function of Z is ⇥ ⇤ ⇥ ⇤ 1 ⇥ ⇤ 1 ⇥ ⇤ b (t). E eitZ = E eitXY = E eitX + E e-itX = E [cos(tX)] = Re µ 2 2 Therefore, Reb µ is a characteristic function. Let X 0 be independent from X and have also distribution µ. Then, the characteristic function of X - X 0 is h i ⇥ ⇤ ⇥ ⇤ 0 E eit(X-X ) = E eitX E e-itX = |b µ(t)|2 . Therefore, |b µ|2 is a characteristic function. If Xi has characteristic function ofPµi , and ifQthe Xi ’s are independent, n bi by independence. then the characteristic function of n i=1 Xi is i=1 µ Pn Finally, suppose Y = i=1 Xi 1{Z=i} where Z is independent from the X’s, and P P{Z = k} = 1/n for k = 1, . . . , n. Then, the characteristic function of bi . Y is n i=1 pi µ b (t) is E[cos(tX)] - iE[sin(tX)] = µ b (-t). Finally, note that the conjugate of µ