Section 14

Distribution Functions

[9/33%]
Show
LaTeX
Problem 14.1

The general nondecreasing function FF has at most countably many discontinuities. Prove this by considering the open intervals

(sup⁡u<xF(u),inf⁡ι>xF(v)) \left(\sup _{u<x} F(u), \inf _{\iota >x} F(v)\right)

-each nonempty one contains a rational.

?
Problem 14.2

For distribution functions FF, the second proof of Theorem 14.1 shows how to construct a measure μ\mu on (R1,R1)(R^{1}, \mathscr {R}^{1}) such that μ(a,b]=F(b)−F(a)\mu (a, b]=F(b)-F(a).

?
(a)

Extend to the case of bounded FF.

(b)

Extend to the general case. Hint: Let Fn(x)F_{n}(x) be −n-n or F(x)F(x) or nn as F(x)<−nF(x)<-n or −n≤F(x)<n-n \leq F(x)<n or n≤F(x)n \leq F(x). Construct the corresponding μn\mu_{n} and define μ(A)=lim⁡nμn(A)\mu (A)=\lim_{n} \mu_{n}(A).

Problem 14.3
?
(a)

Suppose that XX has a continuous, strictly increasing distribution function FF. Show that the random variable F(X)F(X) is uniformly distributed over the unit interval in the sense that P(F(X)≤u)=u\mathbb {P}\left(F(X) \leq u\right)=u for 0≤u≤10 \leq u \leq 1. Passing from XX to F(X)F(X) is called the probability transformation.

(b)

Show that the function φ(u)\varphi (u) defined by (14.5) satisfies F(φ(u)−)≤u≤F(φ(u))F(\varphi (u)-) \leq u \leq F(\varphi (u)) and that, if FF is continuous (but not necessarily strictly increasing), then F(φ(u))=uF(\varphi (u))=u for 0<u<10<u<1.

(c)

Show that P(F(X)<u)=F(φ(u)−)\mathbb {P}\left(F(X)<u\right)=F(\varphi (u)-) and hence that the result in part (a) holds as long as FF is continuous.

†{ }^{\dagger } This theory is associated with the names of Fisher, Fréchet, Gnedenko, and Tippet. For further information, see Galambos.

Problem 14.4

↑ Let CC be the set of continuity points of FF.

?
(a)

Show that for every Borel set A,P(F(X)∈A,X∈C)A, \mathbb {P}\left(F(X) \in A, X \in C\right) is at most the Lebesgue measure of AA.

(b)

Show that if FF is continuous at each point of F−1AF^{-1} A, then P(F(X)∈A)\mathbb {P}\left(F(X) \in A\right) is at most the Lebesgue measure of AA.

Problem 14.5

The Lévy distance d(F,G)d(F, G) between two distribution functions is the infimum of those ϵ\epsilon such that G(x−ϵ)−ϵ≤F(x)≤G(x+ϵ)+ϵG(x-\epsilon )-\epsilon \leq F(x) \leq G(x+\epsilon )+\epsilon for all xx. Verify that this is a metric on the set of distribution functions. Show that a necessary and sufficient condition for Fn⇒FF_{n} \Rightarrow F is that d(Fn,F)→0d\left(F_{n}, F\right) \rightarrow 0.

?
Problem 14.6

12.3 ↑ A Borel function satisfying Cauchy's equation [A20] is automatically bounded in some interval and hence satisfies f(x)=xf(1)f(x)=x f(1). Hint: Take KK large enough that λ[x:x>s,∣f(x)∣≤K]>0\lambda [x: x>s,\left|f(x)\right| \leq K]>0. Apply Problem 12.3 and conclude that ff is bounded in some interval to the right of 0

?
Problem 14.7

↑ Consider sets SS of reals that are !inearly independent over the field of rationals in the sense that n1x1+⋯+nkxk=0n_{1} x_{1}+\cdots +n_{k} x_{k}=0 for distinct points xix_{i} in SS and integers nin_{i} (positive or negative) is impossible unless ni≡0n_{i} \equiv 0.

?
(a)

By Zorn's lemma find a maximal such SS. Show that it is a Hamel basis. That is, show that each real xx can be written uniquely as x=n1x1+⋯+nkxkx=n_{1} x_{1}+\cdots +n_{k} x_{k} for distinct points xix_{i} in SS and integers nin_{i}.

(b)

Define ff arbitrarily on SS, and define it elsewhere by f(n1x1+⋯+nkxk)=n1f(x1)+⋯+nkf(xk)f\left(n_{1} x_{1}+\cdots +n_{k} x_{k}\right) =n_{1} f\left(x_{1}\right)+\cdots +n_{k} f\left(x_{k}\right). Show that ff satisfies Cauchy's equation but need not satisfy f(x)=xf(1)f(x)=x f(1).

(c)

By means of Problem 14.6 give a new construction of a nonmeasurable function and a nonmeasurable set.

Problem 14.8
?
(a)

14.5↑14.5 \uparrow Show that if a distribution function FF is everywhere continuous, then it is uniformly continuous.

(b)

Let δF(ϵ)=sup⁡[F(x)−F(y):∣x−y∣≤ϵ]\delta_{F}(\epsilon )=\sup [F(x)-F(y):\left|x-y\right| \leq \epsilon ] be the modulus of continuity of FF. Show that d(F,G)<ϵd(F, G)<\epsilon implies that sup⁡x∣F(x)−G(x)∣≤ϵ+δF(ϵ)\sup_{x}\left|F(x)-G(x)\right| \leq \epsilon +\delta_{F}(\epsilon ).

(c)

Show that, if Fn⇒FF_{n} \Rightarrow F and FF is everywhere continuous, then Fn(x)→F(x)F_{n}(x) \rightarrow F(x) uniformly in xx. What if FF is continuous over a closed interval?

Problem 14.9

Show that (14.24) and (14.25) are everywhere infinitely differentiable, although not analytic.

?