06.06.2013 Views

Theory of Statistics - George Mason University

Theory of Statistics - George Mason University

Theory of Statistics - George Mason University

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

726 0 Statistical Mathematics<br />

Pro<strong>of</strong>.<br />

Assume the hypothesis: that is, fn ↗ f for all ω ∈ A where ν(Ac ) =<br />

0. Now restrict each function to A, and observe that fnIA ↗ fIA and<br />

fnIA dν = fn dν and fIA dν = f dν. Lemma 0.1.23.1 immediately<br />

implies fn dν ↗ f dν.<br />

That Theorem 0.1.23 follows so readily from Lemma 0.1.23.1 is another<br />

illustration <strong>of</strong> a principle <strong>of</strong> real analysis stated by Littlewood that every convergent<br />

sequence <strong>of</strong> functions is “nearly” uniformly convergent (see page 753).<br />

In the hypotheses <strong>of</strong> the lemma, we have only pointwise convergence. Without<br />

needing uniform convergence, however, we extend the conclusion to the case<br />

<strong>of</strong> convergence a.e.<br />

Theorem 0.1.24 (Fatou’s lemma)<br />

For nonnegative integrable Borel fn,<br />

<br />

<br />

liminf fn dν ≤ liminf n n<br />

fn dν. (0.1.49)<br />

Pro<strong>of</strong>.<br />

Let gn = infk≥n fk and g = limn inf fn. As in the monotone convergence<br />

theorem, gn is nonnegative and gn ↗ g, so gn dν ↗ g dν. Also, for each<br />

n, fn dν ≥ gn dν; hence, we have the desired conclusion.<br />

The next theorem is the most powerful <strong>of</strong> the convergence theorems for<br />

integrals.<br />

Theorem 0.1.25 (Lebesgue’s dominated convergence)<br />

If limn→∞ fn = f a.e. and there exists an integrable function g such that<br />

|fn| ≤ g a.e., then<br />

Pro<strong>of</strong>.<br />

***<br />

<br />

lim<br />

n→∞<br />

<br />

fn dν = f dν. (0.1.50)<br />

Corollary 0.1.25.1 (bounded convergence)<br />

Let {fn} be a sequence <strong>of</strong> measurable functions defined on a set A, where<br />

ν(A) < ∞. If for some real number M, |fn(ω)| ≤ M, and limn→∞ fn(ω) =<br />

f(ω) for each ω ∈ A then<br />

<br />

lim fn dν = f dν. (0.1.51)<br />

n→∞<br />

A A<br />

0.1.6.5 Integrals over More than One Measure Space<br />

So far the integrals we have discussed have been for functions over a single<br />

measure space. We now consider integrals over more than one measure space.<br />

<strong>Theory</strong> <strong>of</strong> <strong>Statistics</strong> c○2000–2013 James E. Gentle

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!