Prof. N. Kajino, Probability Theory WS 2012/2013
Problem set 5, submit solutions by 17.10.2012
TheProblems below will be discussed in the tutorial on 19.10.2012.
(TheExercises are additional and will be discussed only if time permits.) Throughout this problem set,.X;M; /denotes a given measure space.
Problem 1.26([7, Chapter 1, Exercise 9]). Let˛2.0;1/, letf WX !Œ0;1beM- measurable and supposeR
Xf d2.0;1/. Find the limit (with log1 WD 1˛WD 1)
nlim!1
Z
X
nlog 1C.f =n/˛ d:
(Note that log.1Cx/ 0for anyx 2 Œ0;1. If˛ 1, then log.1Cx˛/˛x for anyx2Œ0;1. If˛ < 1, then Fatou’s lemma (Theorem 1.27) applies.)
Exercise 1.27(30 points). Letf W X ! Œ 1;1. Prove that the following three conditions are equivalent:
(1)f isM-measurable.
(2) There existM-measurable functionsf1; f2WX !Œ 1;1such thatf1 f f2onX andf1Df2-a.e.
(3) There exists aM-measurable functionf0WX !Œ 1;1such thatf0Df -a.e.
((1) ) (2): Iff isM-simple, then (2) can be proved by using the definition of M. For a generalM-measurablef, take non-decreasing sequences¹sn˙º1nD1ofM- simple functions converging tof˙, use (2) forsn˙and take lim supn!1, lim infn!1.) Problem 1.28. Letp2.0;1/and letf 2Lp./. Prove that
nlim!1
Z
X
ˇˇf f1¹jfjnº
ˇˇpdD0: (1.81)
Problem 1.29. Letp; q2.0;1/,p < q, and letf WX !Œ0;1beM-measurable.
Prove that
Z
X
fpd 1=p
Z
X
fqd 1=q
.X /.q p/=pq: (1.82) By Problem 1.29, if.X / <1, thenLq.X; /Lp.X; /for anyp; q2.0;1/ withp < q.
Problem 1.30(Minkowski’s inequality). Letp2Œ1;1/and letf; gWX !Œ0;1be M-measurable. Prove that
Z
X
.f Cg/pd 1=p
Z
X
fpd 1=p
C Z
X
gpd 1=p
: (1.83)
(Assumingp > 1, apply H¨older’s inequality tof .f Cg/p 1andg.f Cg/p 1.) For the next problem, we need the following definition.
9
Definition. Letf WX !RandfnWX !R,n2N, beM-measurable. We say that
¹fnº1nD1converges in-measure tof if and only if for any"2.0;1/,
nlim!1 ¹x2X j jfn.x/ f .x/j "º
D0: (1.84)
Problem 1.31. Letf WX !RandfnWX!R,n2N, beM-measurable.
(1) Let p 2 .0;1/and suppose limn!1kfn fkLp./ D 0. Prove that¹fnº1nD1
converges in-measure tof. (Use Problem 1.20-(2) with'.x/Dxp,x2Œ0;1.) (2) Suppose that¹fnº1nD1converges in-measure tof. Prove that there exists a strictly increasing sequence¹nkº1kD1Nsuch that limk!1fnk.x/Df .x/for-a.e.x2X.
(Choosenk 2Nso that ¹x 2X j jfnk.x/ f .x/j 2 kº
2 kand use Problem 1.11-(2) to show that Xnlim infk!1¹x2Xj jfnk.x/ f .x/j< 2 kº
D0.) Problem 1.32. LetA2 M, and define a measurejAonMjA D ¹B\Aj B 2Mº byjA WDjMjA (note thatMjA M). Letf W X !Œ 1;1beM-measurable.
Prove thatR
Xf1Adexists if and only ifR
AfjAd.jA/exists, and in this case Z
A
f dWD Z
X
f1AdD Z
A
fjAd.jA/: (1.85) (Modify the proof of Theorem 1.43. It suffices to prove (1.85) forf˙and hence when f 0. Show first for non-negativeM-simple functions by using Proposition 1.25 and then use Proposition 1.19 and Theorem 1.24 for general non-negativef.)
According to Problem 1.32,R
Af dcould alternatively be defined as the integral offjAwith respect tojADjMjA, therestriction oftoA.
Exercise 1.33. LetN be a-algebra in X such that N M, and let f W X ! Œ 1;1beN-measurable. Prove thatR
Xf dexists if and only ifR
Xf d.jN/exists (note thatjNis a measure on.X;N/), and in this case
Z
X
f dD Z
X
f d.jN/: (1.86)
Exercise 1.34. Letf W X !Œ0;1beM-measurable and-integrable. Prove that, for any"2.0;1/there existsı 2.0;1/such thatR
Af d < "for anyA2Mwith .A/ < ı. (Proof by contradiction. Problem 1.11-(2) can be used.)
10