Homepage Solution manuals Terence Tao Probability Theory Exercise 0.24 (Extended-real measurable functions)

Exercise 0.24 (Extended-real measurable functions)

Let (Ω,) be a measurable space and (,) be the extended reals with the Borel σ-algebra. Let f,g : Ω [,+] be functions taking values in the extended reals.

1.
Show that f is measurable if and only if the sets {ω Ω : f(ω) t} are measurable for all reals t.
2.
Show that f = g if and only if {ω Ω : f(ω) t} = {ω Ω : g(ω) t} for all reals t.
3.
If f1,f2, : Ω [,+] are measurable, show that sup nfn, inf nfn, limsup nfn and liminf nfn are all measurable.

Answers

Suppose that f is measurable. Since any closed set is measurable, then the inverse image f1([,t]) of the closed set [,t] under the measurable function f must be measurable in (Ω,B).
[] We use the measurable induction (Remark 1.4.15 from Measure Theory) with the property P(S) = f1(S) F. We use the open sets of as a generating set for B. Let O O(). Recall that all open subsets of the extended real line can be written as countable unions of open intervals. Any open interval of the extended real line has either one of the following forms:
  • [,a) for a
    The inverse image under f is measurable since

    f1 ([,a)) = f1 ( n=1 [,a 1 n ]) = n=1f1 ( [,a 1 n ])

    and the countable union of measurable sets (by assumption) is measurable.

  • (b,+] for b .
    The inverse image under f is measurable since f1 ((b,+]) = f1 ([,b]) = f1 () f1 ([,b]). The last term is measurable since by the theorem assumption, it is a set difference of two measurable sets.
  • (a,b) for a,b
    This follows by representing (a,b) as (a,+) [,b).

Since any open set is a countable union of the above, it’s inverse image must be measurable as well. In other words, for all

O O() : f1(O) F.

Finally, we demonstrate that the above is a σ-algebra property:

(a)
P() = f1() F is true, since f1() = and the empty set is always contained in any σ-algebra F.
(b)
Suppose that f1(S) F is true for some S Y . Using the set-theoretic property from the Exercise 3.4.4 of Analysis I we see that f1(S) = f1()f1(S) = Ω f1(S). Since F is a σ-algebra that contains f1(S), it must also contain its complement.
(c)
Let S1,S2, X be such that P(Si) = f1(Si) F is true for all i . Then, we also have by Exercise 3.4.3 of Analysis I that f1 ( i=1Si) = i=1f1 (Si) F is true as well since F contains the countable unions of any of its elements.

Hence, we have S B = O() : f1(S) F.

This is obvious.
[] Consider an arbitrary ω¯ from the set {ω Ω : f(ω)g(ω)}. For this ω¯ we have either f(ω¯) < g(ω¯) or f(ω¯) > g(ω¯). Without the loss of generality, consider the first case. Define f(ω¯) < t := f(ω¯) + g(ω¯) 2 < g(ω¯).

We then see that

ω¯ {ω Ω : f(ω) < t}ω¯ {ω Ω : g(ω) < t}

which is a contradiction to the theorem assumption. In other words, {ω Ω : f(ω)g(ω)} = .

1.
We now consider the limiting functions.
  • f(x)sup {fn(x) : n }
    Let E B be a measurable set. Consider f1([,t]) for some arbitrary t . Then it is easy to verify using logical operators that

    f1 ([,t]) = n=1f n1 ([,t]).

    [⊆] Let ω f1 ([,t]). Then n : fn(ω) f(ω) t by definition, and so ω fn1 ([,t]) for all n .
    [⊇] Let ω fn1 ([,t]). Then n : fn(ω) t and so sup nfn(ω) t as well.
    Since a countable union of measurable function is again measurable, the set f1 ([,t]) must be measurable as well.

  • f(x)inf {fn(x) : n }similarly by observing that

    f1 ([,t]) = n=1f n1 ([,t]).

  • f(x)limsup {fn(x) : n }Definition 6.4.6 from Analysis I we can write limsup as

    inf N=msup n=Nf n(x).

    In other words, limsup is a combination of inf and sup and so must be measurable by the previous parts.

  • f(x)liminf {fn(x) : n }
    Follows similarly.
User profile picture
2021-08-01 00:00
Comments