Probability Measures
[23/57%]Define , and for a collection define ; define and . Prove that , and , in the sense that there is equality at each point of . Show that if and only if pointwise. Check the equation and deduce the distributive law See Problem 2.22. . By similar arguments prove that
Let be arbitrary events, and put and , where the union and intersection extend over all the -tuples satisfying . Show that .
Suppose that and that implies . Show that is a field.
Suppose that and that is closed under the formation of complements and finite disjoint unions. Show that need not be a field.
Let be classes of sets in a common space .
Suppose that are fields satisfying . Show that is a field.
Suppose that are -fields satisfying . Show by example that need not be a -field.
The field generated by a class in is defined as the intersection of all fields in containing .
Show that is indeed a field, that , and that is minimal in the sense that if is a field and , then .
Show that for nonempty is the class of sets of the form , where for each and either or , and where the sets , are disjoint. The sets in can thus be explicitly presented, which is not in general true of the sets in .
Show that if consists of the singletons, then is the field in Example 2.3.
Show that , that if is finite, and that .
Show that if is countable, then is countable.
Show for fields and that consists of the finite disjoint unions of sets with . Extend.
Let be a a set lying outside , where is a field [or -field]. Show that the field [or -field] generated by consists of sets of the form
Suppose for each in that is a countable union of elements of . The class of intervals in has this property. Show that coincides with the smallest class over that is closed under the formation of countable unions and intersections.
Show that, if , then there exists a countable subclass of such that .
Show that if contains every subset of , then for each pair and of distinct points in there is in an such that
Show that the reverse implication holds if is countable.
Show by example that the reverse implication need not hold for uncountable
A -field is countably generated, or separable, if it is generated by some countable class of sets.
Show that the -field of Borel sets is countably generated.
Show that the -field of Example 2.4 is countably generated if and only if is countable.
Suppose that and are -fields, , and is countably generated. Show by example that may not be countably generated.
Show that a -field cannot be countably infinite-its cardinality must be finite or else at least that of the continuum. Show by example that a field can be countably infinite.
Let be the field consisting of the finite and the cofinite sets in an infinite , and define on by taking to be 0 or 1 as is finite or cofinite. (Note that is not well defined if is finite.) Show that is finitely additive.
Show that this is not countably additive if is countably infinite.
Show that this is countably additive if is uncountable.
Now let be the -field consisting of the countable and the cocountable sets in an uncountable , and define on by taking to be 0 or 1 as is countable or cocountable. (Note that is not well defined if is countable.) Show that is countably additive.
In let be the class of sets that either (i) are of the first category [A15] or (ii) have complement of the first category. Show that is a -field. For in , take to be 0 in case (i) and 1 in case (ii). Show that is countably additive.
On the field in define to be 1 or 0 according as there does or does not exist some positive (depending on ) such that contains the interval . Show that is finitely but not countably additive. No such example is possible for the field in (Theorem 2.3).
Suppose that is a probability measure on a field . Suppose that for , that for , and that . Extend Theorem 2.1(i) by showing that as . Show that necessarily lies in if it is a -field.
Extend Theorem 2.1(ii) in the same way.
Suppose that is a probability measure on a field , that , and lie in , and that the are nearly disjoint in the sense that for . Show that .
Stochastic arithmetic. Define a set function on the class of all subsets of by
among the first integers, the proportion that lie in is just . Then is a discrete probability measure. The set has density
provided this limit exists. Let be the class of sets having density.
Show that is finitely but not countably additive on .
Show that contains the empty set and and is closed under the formation of complements, proper differences, and finite disjoint unions, but is not closed under the formation of countable disjoint unions or of finite unions that are not disjoint.
Let consist of the periodic sets . Observe that
Show that the field generated by (see Problem 2.5) is contained in . Show that is completely determined on by the value it gives for each to the event that is divisible by .
Assume that diverges (sum over all primes; see Problem 5.20(e)) and prove that , although finitely additive, is not countably additive on the field .
Euler's function is the number of positive integers less than and relatively prime to it. Let be the distinct prime factors of ; from the inclusion-exclusion formula for the events [ ], (2.36), and the fact that the divide , deduce
Show for that for some .
Show that is translation invariant: If , then has a density if and only if does, in which case .
A probability measure space is nonatomic if implies that there exists a such that and ( and in , of course).
Assuming the existence of Lebesgue measure on , prove that it is nonatomic.
Show in the nonatomic case that and imply that there exists a such that and .
Show in the nonatomic case that implies that there exists a such that and . Hint: Inductively define classes , numbers , and sets by , , and . Consider .
Show in the nonatomic case that, if are nonnegative and add to 1, then can be decomposed into sets such that .
Generalize the construction of product measure: For , let be a finite space with given probabilities Let be the space of sequences (2.15), where now . Define on the class of cylinders, appropriately defined, by using the product , on the right in (2.21) Prove countably additive on , and extend Theorem 2.3 and its lemma to this more general setting. Show that the lemma fails if any of the are infinite
Suppose that is a countable partition of Show (see (2.27)) that coincides with . This is a case where can be constructed "from the inside."
Show that the set of normal numbers lies in .
Show that if and only if is a -field. Show that is strictly smaller than for all .
Extend (2.27) to infinite ordinals by defining . Show that, if is the first uncountable ordinal, then . Show that, if the cardinality of does not exceed that of the continuum, then the same is true of . Thus has the power of the continuum.
Extend (2.29) to ordinals as follows. Replace the right side of (2.28) by . Suppose that is defined for Let be a sequence of ordinals such that and such that if , then for infinitely many even and for infinitely many odd ; define
Prove by transfinite induction that (2.38) is in if the are, that every element of has the form (2.38) for sets in , and that (2.31) holds with in place of . Define , and show that : lies in for . Show that is strictly smaller than for .