Q
stringlengths
4
3.96k
A
stringlengths
1
3k
Result
stringclasses
4 values
Theorem 10. If \( A\left( z\right) = \mathop{\sum }\limits_{{n = 0}}^{\infty }{a}_{n}{z}^{n} \), where \( {a}_{n} \geq 0 \) for all \( n \) and the series converges for \( 0 \leq z < 1 \), then we have\n\n\[ \mathop{\lim }\limits_{{n \rightarrow \infty }}\frac{1}{n + 1}\mathop{\sum }\limits_{{v = 0}}^{n}{a}_{v} = \math...
To get a feeling for this theorem, suppose all \( {a}_{n} = c > 0 \) . Then\n\n\[ A\left( z\right) = c\mathop{\sum }\limits_{{n = 0}}^{\infty }{z}^{n} = \frac{c}{1 - z} \]\n\nand the relation in (8.6.5) reduces to the trivial identity\n\n\[ \frac{1}{n + 1}\mathop{\sum }\limits_{{v = 0}}^{n}c = c = \left( {1 - z}\right)...
No
Theorem 11. If \( I \) is finite and forms a single class (namely if there are only a finite number of states and they all communicate with each other), then the chain is necessarily recurrent.
Proof: Suppose the contrary; then each state is transient and so almost surely the particle can spend only a finite number of time units in it by Theorem 7. Since the number of states is finite, the particle can spend altogether only a finite number of time units in the whole space \( I \) . But time increases ad infin...
Yes
Theorem 13. Suppose that we have, for every \( j \) ,\n\n\[ P\left\{ {{X}_{0} = j}\right\} = {w}_{j} \]\n\n(8.6.10)\n\nthen the same is true when \( {X}_{0} \) is replaced by any \( {X}_{n}, n \geq 1 \) . Furthermore the joint probability\n\n\[ P\left\{ {{X}_{n + v} = {j}_{v},0 \leq v \leq l}\right\} \]\n\n(8.6.11)\n\n...
Proof: We have by (8.6.9)\n\n\[ P\left\{ {{X}_{n} = j}\right\} = \mathop{\sum }\limits_{i}P\left\{ {{X}_{0} = i}\right\} {P}_{i}\left\{ {{X}_{n} = j}\right\} = \mathop{\sum }\limits_{i}{w}_{i}{p}_{ij}^{\left( n\right) } = {w}_{j}. \]\n\nSimilarly the probability in (8.6.11) is equal to\n\n\[ P\left\{ {{X}_{n} = {j}_{0}...
Yes
A switch may be on off; call these two positions states 1 and 2. After each unit of time the state may hold or change, but the respective probabilities depend only on the present position. Thus we have a homogeneous Markov chain with \( I = \{ 1,2\} \) and\n\n\[ \n\Pi = \left\lbrack \begin{array}{ll} {p}_{11} & {p}_{12...
Clearly the second equation is just the negative of the first and may be discarded. Solving the first equation we get\n\n\[ \n{x}_{2} = \frac{1 - {p}_{11}}{{p}_{21}}{x}_{1} = \frac{{p}_{12}}{{p}_{21}}{x}_{1} \n\]\n\nThus\n\n\[ \n{w}_{1} = \frac{{p}_{21}}{{p}_{12} + {p}_{21}},\;{w}_{2} = \frac{{p}_{12}}{{p}_{12} + {p}_{...
Yes
Let us find the stationary distribution for the Ehrenfest model.
We can proceed exactly as in Example 15, leading to the formula (8.6.15), but this time it stops at \( j = c \) . Substituting the numerical values from (8.3.16), we obtain\n\n\[ \n{c}_{j} = \frac{c\left( {c - 1}\right) \cdots \left( {c - j + 1}\right) }{1 \cdot 2\cdots j} = \left( \begin{array}{l} c \\ j \end{array}\r...
Yes
Theorem 14. The \( \left\{ {y}_{i}\right\} \) above satisfies the system of equations\n\n\[ \n{x}_{i} = \mathop{\sum }\limits_{{j \in T}}{p}_{ij}{x}_{j} + \mathop{\sum }\limits_{{j \in C}}{p}_{ij},\;i \in T.\n\]\n\n(8.7.3)\n\nIf \( T \) is finite, it is the unique solution of this system. Hence it can be computed by st...
Proof: Let the particle start from \( i \), and consider its state \( j \) after one step. If \( j \in T \), then the Markov property shows that the conditional probability of absorption becomes \( {y}_{j} \) ; if \( j \in C \), then it is already absorbed; if \( j \in \) \( \left( {I - T}\right) - C \), then it can ne...
Yes
Let us return to Problem 1 of \( §{8.1} \), where \( t = c - 1 \) . If \( p \neq q \) , then \( {v}_{i} = {\left( q/p\right) }^{i} \) is a nonconstant solution of (8.7.6).
This is trivial to verify but you may well demand to know how on earth did we discover such a solution? The answer in this case is easy (but motivated by knowledge of difference equations used in \( §{8.1} \) ): try a solution of the form \( {\lambda }^{i} \) and see what \( \lambda \) must be. Now if we substitute thi...
No
The following model of random reproduction was introduced by S. Wright in his genetical studies (see, e.g., [Karlin] for further details). In a haploid organism the genes occur singly rather than in pairs as in the diploid case considered in \( §{5.6} \) . Suppose \( {2N} \) genes of types \( A \) and \( a \) (the alle...
It follows that [see (4.4.16) or (6.3.6)] the expected number of \( A \) genes is equal to\n\n\[ \n\mathop{\sum }\limits_{{j = 0}}^{{2N}}{p}_{ij}j = {2N}\frac{i}{2N} = i \n\]\n\n(8.7.12)\n\nThis means that the expected number of \( A \) genes in the next generation is equal to the actual (but random) number of these ge...
Yes
What is the distribution of the number of particles of the second generation?
Let the generating function of \( {X}_{1} \) be \( g \) :\n\n\[ g\left( z\right) = \mathop{\sum }\limits_{{j = 0}}^{\infty }{a}_{j}{z}^{j} \]\n\nSuppose the number of particles in the first generation is equal to \( j \), and we denote the numbers of their descendants by \( {Z}_{1},\ldots ,{Z}_{j} \), respectively. The...
Yes
Consider \( M = 2 \) assets. Assume that \( {\sigma }^{2}\left( {R}_{1}\right) = {\sigma }^{2}\left( {R}_{2}\right) > 0 \) , i.e., the two assets have the same risk. Assume further that \( - 1 \leq {\rho }_{12} < 1 \) , where \( {\rho }_{12} \) is the correlation coefficient between \( {R}_{1} \) and \( {R}_{2} \) (see...
Indeed,\n\n\[ \n{\sigma }^{2}\left( {\alpha {R}_{1} + \left( {1 - \alpha }\right) {R}_{2}}\right) = {\sigma }^{2}\left( {R}_{1}\right) + {2\alpha }\left( {1 - \alpha }\right) \left\lbrack {\operatorname{Cov}\left( {{R}_{1},{R}_{2}}\right) - \sigma \left( {R}_{1}\right) \sigma \left( {R}_{2}\right) }\right\rbrack \n\]\n...
Yes
In the setting of Example 1, is there an allocation \( \left( {\alpha ,1 - \alpha }\right) \) , with \( 0 < \alpha < 1 \), that is least risky? In other words, is there \( \alpha \in \left( {0,1}\right) \) that minimizes the function \( V\left( \alpha \right) \) defined by\n\n\[ V\left( \alpha \right) = {\sigma }^{2}\l...
From (9.4.1) we see that \( V \) is a quadratic function in \( \alpha \), with second derivative\n\n\[ {V}^{\prime \prime }\left( \alpha \right) = 2{\sigma }^{2}\left( {R}_{1}\right) + 2{\sigma }^{2}\left( {R}_{2}\right) - 4{\rho }_{12}\sigma \left( {R}_{1}\right) \sigma \left( {R}_{2}\right) .\n\nUnder the assumptions...
Yes
Consider again two assets such that \( \sigma \left( {R}_{1}\right) > 0 \) and \( \sigma \left( {R}_{2}\right) = \) 0 . The second asset is a riskless security as we saw previously. Then for \( 0 \leq \alpha \leq 1, V\left( \alpha \right) = {\sigma }^{2}\left( {\alpha {R}_{1} + \left( {1 - \alpha }\right) {R}_{2}}\righ...
Thus the minimand \( {\alpha }^{ * } \) of the return variance of the portfolio \( \left( {\alpha ,1 - \alpha }\right) \) is \( {\alpha }^{ * } = 0 \) and \( V\left( {\alpha }^{ * }\right) = 0 \) is the smallest risk.
Yes
Theorem 4. Let \( \left\{ {X}_{i}\right\} \) be independent, following the same stable distribution with characteristic function \( \varphi \left( \theta \right) = {e}^{-{\gamma }_{\alpha }{\left| \theta \right| }^{\alpha } + {id\theta }} \) . Then, for a given positive \( a,{Y}_{n} = \left( {1/a}\right) {n}^{-1/\alpha...
Proof: For \( 1 \leq k \leq n \), let \( {X}_{k}^{\prime } = \left( {{X}_{k} - d}\right) /\left( {a{n}^{1/\alpha }}\right) \) . Its characteristic function is\n\n\[ \n{\varphi }^{\prime }\left( \theta \right) = E\left\lbrack {e}^{{i\theta }{X}_{k}^{\prime }}\right\rbrack \n\]\n\n\[ \n= E\left\lbrack {e}^{{i\theta }\fra...
Yes
Theorem 1. Under the same conditions of the theorem in Appendix 3, we have for a finite optional time \( T \)\n\n\[ E\left( {X}_{T}\right) \geq E\left( {X}_{0}\right) \text{ for a submartingale }\]\n\n\[ \text{and}\;E\left( {X}_{T}\right) \leq E\left( {X}_{0}\right) \text{for a supermartingale.} \]
Proof: The proof is exactly as in Appendix 3 with equalities replaced with the corresponding inequalities (left as an exercise).
No
The time-0 price of a European option on a stock with price \( S \) and payoff \( g \) at time 1 is\n\n\[ \n{V}_{0} = \frac{1}{1 + r}\left\lbrack {\widetilde{p}g\left( H\right) + \left( {1 - \widetilde{p}}\right) g\left( T\right) }\right\rbrack \n\]\n\nwhere\n\n\[ \n\widetilde{p} = \frac{1 + r - d}{u - d} \n\]
Proof: We first show that \( {V}_{0} = {\beta }_{0}{S}_{0} + {\gamma }_{0}{B}_{0} \), where \( {\beta }_{0} \) and \( {\gamma }_{0} \) are defined by (10.2.2). Suppose \( \epsilon = {V}_{0} - \left( {{\beta }_{0}{S}_{0} + {\gamma }_{0}{B}_{0}}\right) > 0 \) . Consider the following portfolio \( \left( {-1,{\beta }_{0},...
Yes
Proposition 2. The values \( {V}_{0},{V}_{1},\ldots ,{V}_{N} \) of a European option expiring at time \( {t}_{N} \) satisfy the recursive relations\n\n\[ \n{V}_{n - 1}\left( {\widehat{\omega }}_{n - 1}\right) = \frac{1}{1 + r}\left\lbrack {\widetilde{p}{V}_{n}\left( {{\widehat{\omega }}_{n - 1},0}\right) + \left( {1 - ...
Proof: As in the proof of Proposition 1, we need to first show that \( {V}_{n - 1} = \) \( {\beta }_{n - 1}{S}_{n - 1} + {\gamma }_{n - 1}{B}_{n - 1} \), where \( {\beta }_{n - 1} \) and \( {\gamma }_{n - 1} \) are defined in (10.3.1). This step is straightforward and is left as an exercise.
No
Proposition 3. The time-0 price of a European call option on a stock with price process \( S \) defined above and payoff \( g \) at time \( N \) is\n\n\[ \n{V}_{0} = \frac{1}{{\left( 1 + r\right) }^{N}}\widetilde{E}\left\{ {g \mid \left( {{S}_{0},{B}_{0}}\right) }\right\} \]\n\nwhere \( \widetilde{E} \) refers to the e...
Proof: As in the one-period case (cf. Proposition 1) we can write, for \( 1 \leq i \leq N \)\n\n\[ \n{V}_{i - 1} = E\left\{ {\frac{1}{1 + r}{V}_{i} \mid \left( {{S}_{i - 1},{B}_{i - 1}}\right) }\right\} .\n\nThen\n\n\[ \n{V}_{i - 2} = E\left\{ {\frac{1}{1 + r}{V}_{i - 1} \mid \left( {{S}_{i - 2},{B}_{i - 2}}\right) }\r...
No
Consider now\n\n(5)\n\[ f\left( x\right) = \mathop{\sum }\limits_{{n = 1}}^{\infty }{b}_{n}{\delta }_{{a}_{n}}\left( x\right) \]
Since \( 0 \leq {\delta }_{{a}_{n}}\left( x\right) \leq 1 \) for every \( n \) and \( x \), the series in (5) is absolutely and uniformly convergent. Since each \( {\delta }_{{a}_{n}} \) is increasing, it follows that if \( {x}_{1} < {x}_{2} \), \n\n\[ f\left( {x}_{2}\right) - f\left( {x}_{1}\right) = \mathop{\sum }\li...
Yes
Theorem 1.2.1. Let\n\n\[ \n{F}_{c}\left( x\right) = F\left( x\right) - {F}_{d}\left( x\right) \]\n\nthen \( {F}_{c} \) is positive, increasing, and continuous.
PROOF. Let \( x < {x}^{\prime } \), then we have\n\n\[ \n\text{4)}\;{F}_{d}\left( {x}^{\prime }\right) - {F}_{d}\left( x\right) = \mathop{\sum }\limits_{{x < {a}_{j} \leq {x}^{\prime }}}{b}_{j} = \mathop{\sum }\limits_{{x < {a}_{j} \leq {x}^{\prime }}}\left\lbrack {F\left( {a}_{j}\right) - F\left( {{a}_{j} - }\right) }...
Yes
Theorem 1.2.2. Let \( F \) be a d.f. Suppose that there exist a continuous function \( {G}_{c} \) and a function \( {G}_{d} \) of the form\n\n\[ \n{G}_{d}\left( x\right) = \mathop{\sum }\limits_{j}{b}_{j}^{\prime }{\delta }_{{a}_{j}^{\prime }}\left( x\right)\n\]\n\n[where \( \left\{ {a}_{j}^{\prime }\right\} \) is a co...
PROOF. If \( {F}_{d} \neq {G}_{d} \), then either the sets \( \left\{ {a}_{j}\right\} \) and \( \left\{ {a}_{j}^{\prime }\right\} \) are not identical, or we may relabel the \( {a}_{j}^{\prime } \) so that \( {a}_{j}^{\prime } = {a}_{j} \) for all \( j \) but \( {b}_{j}^{\prime } \neq {b}_{j} \) for some \( j \) . In e...
Yes
Theorem 2.1.1. A field is a B.F. if and only if it is also an M.C.
PROOF. The \
No
Theorem 2.1.2. Let \( {\mathcal{T}}_{0} \) be a field, \( \mathcal{G} \) the minimal M.C. containing \( {\mathcal{T}}_{0},\mathcal{F} \) the minimal B.F. containing \( {\mathcal{F}}_{0} \), then \( \mathcal{F} = \mathcal{G} \) .
PROOF. Since a B.F. is an M.C., we have \( \mathcal{F} \supset \mathcal{G} \) . To prove \( \mathcal{F} \subset \mathcal{G} \) it is sufficient to show that \( \mathcal{G} \) is a B.F. Hence by Theorem 2.1.1 it is sufficient to show that \( \mathcal{G} \) is a field. We shall show that it is closed under intersection a...
Yes
Theorem 2.2.1. The axioms of finite additivity and of continuity together are equivalent to the axiom of countable additivity.
PROOF. Let \( {E}_{n} \downarrow \) . We have the obvious identity:\n\n\[ \n{E}_{n} = \mathop{\bigcup }\limits_{{k = n}}^{\infty }\left( {{E}_{k} \smallsetminus {E}_{k + 1}}\right) \cup \mathop{\bigcap }\limits_{{k = 1}}^{\infty }{E}_{k} \n\]\n\nIf \( {E}_{n} \downarrow \varnothing \), the last term is the empty set. H...
Yes
Let \( \Omega \) be a countable set: \( \Omega = \left\{ {{\omega }_{j}, j \in J}\right\} \), where \( J \) is a countable index set, and let \( \mathcal{F} \) be the total B.F. of \( \Omega \) . Choose any sequence of numbers \( \left\{ {{p}_{j}, j \in J}\right\} \) satisfying\n\n\[ \forall j \in J : {p}_{j} \geq 0;\;...
In words, we assign \( {p}_{i} \) as the value of the \
No
Example 2. Let \( \mathcal{U} = (0,1\rbrack ,\ell \) the collection of intervals:\n\n\[ \mathcal{C} = \{ (a, b\rbrack : 0 < a < b \leq 1\}\]\n\n\( \mathcal{B} \) the minimal B.F. containing \( \mathcal{C}, m \) the Borel-Lebesgue measure on \( \mathcal{B} \) . Then \( \left( {\mathcal{U},\mathcal{B}, m}\right) \) is a ...
Let \( {\mathcal{B}}_{0} \) be the collection of subsets of \( \mathcal{U} \) each of which is the union of a finite number of members of \( \ell \) . Thus a typical set \( B \) in \( {\mathcal{B}}_{0} \) is of the form\n\n\[ B = \mathop{\bigcup }\limits_{{j = 1}}^{n}\left( {{a}_{j},{b}_{j}}\right\rbrack \;\text{ where...
No
Theorem 2.2.3. Let \( \mu \) and \( v \) be two measures defined on the same B.F. \( \mathcal{F} \) , which is generated by the field \( {\mathcal{T}}_{0} \). If either \( \mu \) or \( v \) is \( \sigma \) -finite on \( {\mathcal{T}}_{0} \), and \( \mu \left( E\right) = v\left( E\right) \) for every \( E \in {\mathcal{...
PROOF. We give the proof only in the case where \( \mu \) and \( v \) are both finite, leaving the rest as an exercise. Let\n\n\[ \mathcal{C} = \{ E \in \mathcal{F} : \mu \left( E\right) = v\left( E\right) \}\]\n\nthen \( \ell \supset {\mathcal{T}}_{0} \) by hypothesis. But \( \ell \) is also a monotone class, for if \...
No
Theorem 3.1.1. For any function \( X \) from \( \Omega \) to \( {\mathcal{R}}^{1} \) (or \( {\mathcal{R}}^{ * } \) ), not necessarily an r.v., the inverse mapping \( {X}^{-1} \) has the following properties:
\[ {X}^{-1}\left( {A}^{c}\right) = {\left( {X}^{-1}\left( A\right) \right) }^{c} \] \[ {X}^{-1}\left( {\mathop{\bigcup }\limits_{\alpha }{A}_{\alpha }}\right) = \mathop{\bigcup }\limits_{\alpha }{X}^{-1}\left( {A}_{\alpha }\right) \] \[ {X}^{-1}\left( {\mathop{\bigcap }\limits_{\alpha }{A}_{\alpha }}\right) = \mathop{\...
Yes
Theorem 3.1.2. \( X \) is an r.v. if and only if for each real number \( x \), or each real number \( x \) in a dense subset of \( {\mathcal{R}}^{1} \), we have\n\n\[ \n\{ \omega : X\left( \omega \right) \leq x\} \in \mathcal{F}.\n\]
PROOF. The preceding condition may be written as\n\n(3)\n\n\[ \n\forall x : {X}^{-1}(\left( {-\infty, x\rbrack }\right) \in \mathcal{F}.\n\]\n\nConsider the collection \( \mathcal{A} \) of all subsets \( S \) of \( {\mathcal{R}}^{1} \) for which \( {X}^{-1}\left( S\right) \in \mathcal{F} \) . From Theorem 3.1.1 and the...
Yes
Theorem 3.1.3. Each r.v. on the probability space \( \left( {\Omega ,\mathcal{F},\mathcal{P}}\right) \) induces a probability space \( \left( {{\mathcal{R}}^{1},{\mathcal{B}}^{1},\mu }\right) \) by means of the following correspondence:\n\n(4)\n\n\[ \forall B \in {\mathcal{B}}^{1} : \;\mu \left( B\right) = \mathcal{P}\...
PROOF. Clearly \( \mu \left( B\right) \geq 0 \) . If the \( {B}_{n} \) ’s are disjoint sets in \( {\mathcal{B}}^{1} \), then the \( {X}^{-1}\left( {B}_{n}\right) \) ’s are disjoint by Theorem 3.1.1. Hence\n\n\[ \mu \left( {\mathop{\bigcup }\limits_{n}{B}_{n}}\right) = \mathcal{P}\left( {{X}^{-1}\left( {\mathop{\bigcup ...
Yes
In this case an r.v. is by definition just a Borel measurable function. According to the usual definition, \( f \) on \( \mathcal{U} \) is Borel measurable iff \( {f}^{-1}\left( {\mathcal{B}}^{1}\right) \subset \mathcal{B} \) .
In particular, the function \( f \) given by \( f\left( \omega \right) \equiv \omega \) is an r.v. The two r.v.’s \( \omega \) and \( 1 - \omega \) are not identical but are identically distributed; in fact their common distribution is the underlying measure \( m \) .
Yes
Theorem 3.1.4. If \( X \) is an r.v., \( f \) a Borel measurable function [on \( \left( {{\mathcal{R}}^{1},{\mathcal{B}}^{1}}\right) \) ], then \( f\left( X\right) \) is an r.v.
PROOF. The quickest proof is as follows. Regarding the function \( f\left( X\right) \) of \( \omega \) as the \
No
Theorem 3.1.5. If \( X \) and \( Y \) are r.v.’s and \( f \) is a Borel measurable function of two variables, then \( f\left( {X, Y}\right) \) is an r.v.
\[ {\left\lbrack f \circ \left( X, Y\right) \right\rbrack }^{-1}\left( {\mathcal{B}}^{1}\right) = {\left( X, Y\right) }^{-1} \circ {f}^{-1}\left( {\mathcal{B}}^{1}\right) \subset {\left( X, Y\right) }^{-1}\left( {\mathcal{B}}^{2}\right) \subset \mathcal{F}. \] The last inclusion says the inverse mapping \( {\left( X, Y...
Yes
Theorem 3.1.6. If \( \left\{ {{X}_{j}, j \geq 1}\right\} \) is a sequence of r.v.’s, then\n\n\[ \mathop{\inf }\limits_{j}{X}_{j},\;\mathop{\sup }\limits_{j}{X}_{j},\;\mathop{\liminf }\limits_{j}{X}_{j},\;\mathop{\limsup }\limits_{j}{X}_{j} \]\n\nare r.v.'s, not necessarily finite-valued with probability one though ever...
PROOF. To see, for example, that \( \mathop{\sup }\limits_{j}{X}_{j} \) is an r.v., we need only observe\n\nthe relation\n\n\[ \forall x \in {\mathcal{R}}^{1} : \left\{ {\mathop{\sup }\limits_{j}{X}_{j} \leq x}\right\} = \mathop{\bigcap }\limits_{j}\left\{ {{X}_{j} \leq x}\right\} \]\n\nand use Theorem 3.1.2. Since\n\n...
Yes
Theorem 3.2.1. We have\n\n\[ \mathop{\sum }\limits_{{n = 1}}^{\infty }\mathcal{P}\left( {\left| X\right| \geq n}\right) \leq \mathcal{E}\left( \left| X\right| \right) \leq 1 + \mathop{\sum }\limits_{{n = 1}}^{\infty }\mathcal{P}\left( {\left| X\right| \geq n}\right) \]\n\nso that \( \mathcal{E}\left( \left| X\right| \r...
PROOF. By the additivity property (iii), if \( {\Lambda }_{n} = \{ n \leq \left| X\right| < n + 1\} \), \n\n\[ \mathcal{E}\left( \left| X\right| \right) = \mathop{\sum }\limits_{{n = 0}}^{\infty }{\int }_{{\Lambda }_{n}}\left| X\right| d\mathcal{P} \]\n\nHence by the mean value theorem (vi) applied to each set \( {\Lam...
Yes
Theorem 3.2.2. Let \( X \) on \( \left( {\Omega ,\mathcal{F},\mathcal{P}}\right) \) induce the probability space \( \left( {{\mathcal{R}}^{1},{\mathcal{B}}^{1},\mu }\right) \) according to Theorem 3.1.3 and let \( f \) be Borel measurable. Then we have\n\n(11)\n\n\[{\int }_{\Omega }f\left( {X\left( \omega \right) }\rig...
PROOF. Let \( B \in {\mathcal{B}}^{1} \), and \( f = {1}_{B} \), then the left side in (11) is \( \mathcal{P}\left( {X \in B}\right) \) and the right side is \( \mu \left( B\right) \) . They are equal by the definition of \( \mu \) in (4) of Sec. 3.1. Now by the linearity of both integrals in (11), it will also hold if...
Yes
Theorem 3.2.3. Let \( \\left( {X, Y}\\right) \) on \( \\left( {\\Omega ,\\mathcal{F},\\mathcal{P}}\\right) \) induce the probability space \( \\left( {{\\mathcal{R}}^{2},{\\mathcal{B}}^{2},{\\mu }^{2}}\\right) \) and let \( f \) be a Borel measurable function of two variables. Then we have\n\n(14)\n\n\[ \n{\\int }_{\\O...
Note that \( f\\left( {X, Y}\\right) \) is an r.v. by Theorem 3.1.5.
No
Theorem 3.3.1. If \( \\left\\{ {{X}_{j},1 \\leq j \\leq n}\\right\\} \) are independent r.v.’s and \( \\left\\{ {{f}_{j},1 \\leq j \\leq }\\right. \) \( n\\} \) are Borel measurable functions, then \( \\left\\{ {{f}_{j}\\left( {X}_{j}\\right) ,1 \\leq j \\leq n}\\right\\} \) are independent r.v.'s.
PROOF. Let \( {A}_{j} \\in {\\mathcal{B}}^{1} \), then \( {f}_{j}^{-1}\\left( {A}_{j}\\right) \\in {\\mathcal{B}}^{1} \) by the definition of a Borel measurable function. By Theorem 3.1.1, we have\n\n\[ \n\\mathop{\\bigcap }\\limits_{{j = 1}}^{n}\\left\\{ {{f}_{j}\\left( {X}_{j}\\right) \\in {A}_{j}}\\right\\} = \\math...
Yes
Theorem 3.3.3. If \( X \) and \( Y \) are independent and both have finite expectations, then\n\n\[ \mathcal{E}\left( {XY}\right) = \mathcal{E}\left( X\right) \mathcal{E}\left( Y\right) \]
PROOF. We give two proofs in detail of this important result to illustrate the methods. Cf. the two proofs of (14) in Sec. 3.2, one indicated there, and one in Exercise 9 of Sec. 3.2.\n\nFirst proof. Suppose first that the two r.v.’s \( X \) and \( Y \) are both discrete belonging respectively to the weighted partition...
Yes
Let \( n \geq 2 \) and \( \left( {{\Omega }_{j},{\mathcal{J}}_{j},{\mathcal{P}}_{j}}\right) \) be \( n \) discrete probability spaces. We define the product space\n\n\[ \n{\Omega }^{n} = {\Omega }_{1} \times \cdots \times {\Omega }_{n}\left( {n\text{ factors }}\right) \n\]\n\nto be the space of all ordered \( n \) -tup...
To see this, we observe that the left side is, by definition, equal to\n\n\[ \n\mathop{\sum }\limits_{{{\omega }_{1} \in {S}_{1}}}\cdots \mathop{\sum }\limits_{{{\omega }_{n} \in {S}_{n}}}{\mathcal{P}}^{n}\left( \left\{ {{\omega }_{1},\ldots ,{\omega }_{n}}\right\} \right) = \mathop{\sum }\limits_{{{\omega }_{1} \in {S...
Yes
Can we construct r.v.’s on the probability space \( \left( {\mathcal{U},\mathcal{B}, m}\right) \) itself, without going to a product space?
Indeed we can, but only by imbedding a product structure in \( \mathcal{U} \). For each real number in \( (0,1\rbrack \), consider its binary digital expansion\n\n\[ x = \cdot {\epsilon }_{1}{\epsilon }_{2}\cdots {\epsilon }_{n}\cdots = \mathop{\sum }\limits_{{n = 1}}^{\infty }\frac{{\epsilon }_{n}}{{2}^{n}},\;\operato...
Yes
Theorem 3.3.4. Let a finite or infinite sequence of p.m.’s \( \left\{ {\mu }_{j}\right\} \) on \( \left( {{\mathcal{R}}^{1},{\mathcal{B}}^{1}}\right) \) , or equivalently their d.f.'s, be given. There exists a probability space \( \left( {\Omega ,\mathcal{F},\mathcal{P}}\right) \) and a sequence of independent r.v.’s \...
PROOF. Without loss of generality we may suppose that the given sequence is infinite. (Why?) For each \( n \), let \( \left( {{\Omega }_{n},{\mathcal{T}}_{n},{\mathcal{P}}_{n}}\right) \) be a probability space\nin which there exists an r.v. \( {X}_{n} \) with \( {\mu }_{n} \) as its p.m. Indeed this is possible if we t...
No
Theorem 3.3.5. Let \( {\mathcal{T}}_{0} \) be a field of subsets of an abstract space \( \Omega \), and \( \mathcal{P} \) a p.m. on \( {\mathcal{T}}_{0} \) . There exists a unique p.m. on the B.F. generated by \( {\mathcal{T}}_{0} \) that agrees with \( \mathcal{P} \) on \( {\mathcal{T}}_{0} \) .
The uniqueness is proved in Theorem 2.2.3.
No
Theorem 3.3.6. For each \( n \geq 1 \), let \( {\mu }^{n} \) be a p.m. on \( \left( {{\mathcal{R}}^{n},{\mathcal{B}}^{n}}\right) \) such that\n\n(16)\n\[ \forall m < n : {\mu }^{n} \circ {\pi }_{mn} = {\mu }^{m}. \]\n\nThen there exists a probability space \( \left( {\Omega ,\mathcal{F},\mathcal{P}}\right) \) and a seq...
For a proof of this first fundamental theorem in the theory of stochastic processes, see Kolmogorov [8].
No
Theorem 4.1.1. The sequence \( \left\{ {X}_{n}\right\} \) converges a.e. to \( X \) if and only if for every \( \epsilon > 0 \) we have\n\n(2)\n\n\[ \mathop{\lim }\limits_{{m \rightarrow \infty }}\mathcal{P}\left\{ {\left| {{X}_{n} - X}\right| \leq \epsilon \text{ for all }n \geq m}\right\} = 1 \]\n\nor equivalently\n\...
PROOF. Suppose there is convergence a.e. and let \( {\Omega }_{0} = \Omega \smallsetminus \mathbf{N} \) where \( \mathbf{N} \) is as in (1). For \( m \geq 1 \) let us denote by \( {A}_{m}\left( \epsilon \right) \) the event exhibited in (2), namely:\n\n(3)\n\n\[ {A}_{m}\left( \epsilon \right) = \mathop{\bigcap }\limits...
Yes
Theorem 4.1.4. If \( {X}_{n} \) converges to 0 in \( {L}^{p} \), then it converges to 0 in pr. The converse is true provided that \( \left\{ {X}_{n}\right\} \) is dominated by some \( Y \) that belongs to \( {L}^{p} \) .
PROOF. By Chebyshev inequality with \( \varphi \left( x\right) \equiv {\left| x\right| }^{p} \), we have\n\n(9)\n\n\[ \mathcal{P}\left\{ {\left| {X}_{n}\right| \geq \epsilon }\right\} \leq \frac{\mathcal{E}\left( {\left| {X}_{n}\right| }^{p}\right) }{{\epsilon }^{p}} \]\n\nLetting \( n \rightarrow \infty \), the right ...
No
Theorem 4.1.5. \( {X}_{n} \rightarrow 0 \) in pr. if and only if\n\n\[ \mathcal{E}\left( \frac{\left| {X}_{n}\right| }{1 + \left| {X}_{n}\right| }\right) \rightarrow 0 \]
PROOF. If \( \rho \left( {X, Y}\right) = 0 \), then \( \mathcal{E}\left( \left| {X - Y}\right| \right) = 0 \), hence \( X = Y \) a.e. by Exercise 1 of Sec. 3.2. To show that \( \rho \left( {\cdot , \cdot }\right) \) is metric it is sufficient to show that\n\n\[ \mathcal{E}\left( \frac{\left| X - Y\right| }{1 + \left| {...
No
Convergence in pr. does not imply convergence in \( {L}^{p} \), and the latter does not imply convergence a.e.
Take the probability space \( \left( {\Omega ,\mathcal{F},\mathcal{P}}\right) \) to be \( \left( {\mathcal{U},\mathcal{B}, m}\right) \) as in Example 2 of Sec. 2.2. Let \( {\varphi }_{k.j} \) be the indicator of the interval\n\n\[ \left( {\frac{j - 1}{k},\frac{j}{k}}\right) ,\;k \geq 1,1 \leq j \leq k.\]\n\nOrder these...
Yes
Example 2. Convergence a.e. does not imply convergence in \( {L}^{p} \) .
In \( \left( {\mathcal{U},\mathcal{B}, m}\right) \) define\n\n\[ \n{X}_{n}\left( \omega \right) = \left\{ \begin{array}{ll} {2}^{n}, & \text{ if }\omega \in \left( {0,\frac{1}{n}}\right) ; \\ 0, & \text{ otherwise. } \end{array}\right. \]\n\nThen \( \mathcal{E}\left( {\left| {X}_{n}\right| }^{p}\right) = {2}^{np}/n \ri...
Yes
Theorem 4.2.1. We have for arbitrary events \( \left\{ {E}_{n}\right\} \) :\n\n(4)\n\[ \mathop{\sum }\limits_{n}\mathcal{P}\left( {E}_{n}\right) < \infty \Rightarrow \mathcal{P}\left( {{E}_{n}\text{ i.o. }}\right) = 0. \]
PROOF. By Boole's inequality for p.m.'s, we have\n\n\[ \mathcal{P}\left( {F}_{m}\right) \leq \mathop{\sum }\limits_{{n = m}}^{\infty }\mathcal{P}\left( {E}_{n}\right) \]\n\nHence the hypothesis in (4) implies that \( \mathcal{P}\left( {F}_{m}\right) \rightarrow 0 \), and the conclusion in (4) now follows by (2).
Yes
Theorem 4.2.2. \( {X}_{n} \rightarrow 0 \) a.e. if and only if\n\n(5)\n\[ \forall \epsilon > 0 : \mathcal{P}\left\{ {\left| {X}_{n}\right| > \epsilon \text{ i.o. }}\right\} = 0. \]
PROOF. Using the notation \( {A}_{m} = \mathop{\bigcap }\limits_{{n = m}}^{\infty }\left\{ {\left| {X}_{n}\right| \leq \epsilon }\right\} \) as in (3) of Sec. 4.1\n\n(with \( X = 0 \) ), we have\n\n\[ \left\{ {\left| {X}_{n}\right| > \epsilon \text{ i.o. }}\right\} = \mathop{\bigcap }\limits_{{m = 1}}^{\infty }\mathop{...
Yes
Theorem 4.2.3. If \( {X}_{n} \rightarrow X \) in pr., then there exists a sequence \( \left\{ {n}_{k}\right\} \) of integers increasing to infinity such that \( {X}_{{n}_{k}} \rightarrow X \) a.e. Briefly stated: convergence in pr. implies convergence a.e. along a subsequence.
PROOF. We may suppose \( X \equiv 0 \) as explained before. Then the hypothesis may be written as\n\n\[ \forall k > 0 : \;\mathop{\lim }\limits_{{n \rightarrow \infty }}\mathcal{P}\left( {\left| {X}_{n}\right| > \frac{1}{{2}^{k}}}\right) = 0. \]\n\nIt follows that for each \( k \) we can find \( {n}_{k} \) such that\n\...
Yes
Theorem 4.2.4. If the events \( \left\{ {E}_{n}\right\} \) are independent, then\n\n(6)\n\[ \mathop{\sum }\limits_{n}\mathcal{P}\left( {E}_{n}\right) = \infty \Rightarrow \mathcal{P}\left( {{E}_{n}\text{ i.o. }}\right) = 1. \]
PROOF. By (3) we have\n\n(7)\n\[ \mathcal{P}\left\{ {\mathop{\liminf }\limits_{n}{E}_{n}^{c}}\right\} = \mathop{\lim }\limits_{{m \rightarrow \infty }}\mathcal{P}\left( {\mathop{\bigcap }\limits_{{n = m}}^{\infty }{E}_{n}^{c}}\right) .\n\nThe events \( \left\{ {E}_{n}^{c}\right\} \) are independent as well as \( \left\...
No
Let \( {X}_{n} = {c}_{n} \) where the \( {c}_{n} \)’s are constants tending to zero. Then \( {X}_{n} \rightarrow 0 \) deterministically.
For any interval \( I \) such that \( 0 \notin \overline{\mathrm{I}} \), where \( \overline{\mathrm{I}} \) is the closure of \( I \), we have \( \mathop{\lim }\limits_{n}{\mu }_{n}\left( I\right) = 0 = \mu \left( I\right) \) ; for any interval such that \( 0 \in {I}^{ \circ } \), where \( {I}^{ \circ } \) is the interi...
Yes
Example 2. Let \( {X}_{n} = {c}_{n} \) where \( {c}_{n} \rightarrow + \infty \) . Then \( {X}_{n} \rightarrow + \infty \) deterministically. According to our definition of a r.v., the constant \( + \infty \) indeed qualifies. But for any finite interval \( \left( {a, b}\right) \) we have \( \mathop{\lim }\limits_{n}{\m...
This example can be easily ramified; e.g. let \( {a}_{n} \rightarrow - \infty \) , \( {b}_{n} \rightarrow + \infty \) and\n\n\[ \n{X}_{n} = \left\{ \begin{array}{ll} {a}_{n} & \text{ with probability }\alpha , \\ 0 & \text{ with probability }1 - \alpha - \beta , \\ {b}_{n} & \text{ with probability }\beta . \end{array}...
Yes
Theorem 4.3.1. Let \( \left\{ {\mu }_{n}\right\} \) and \( \mu \) be s.p.m.’s. The following propositions are equivalent.\n\n(i) For every finite interval \( \left( {a, b}\right) \) and \( \epsilon > 0 \), there exists an \( {n}_{0}\left( {a, b,\epsilon }\right) \) such that if \( n \geq {n}_{0} \), then\n\n\( \mu \lef...
PROOF. To prove that (i) \( \Rightarrow \) (ii), let \( \left( {a, b}\right) \) be a continuity interval of \( \mu \) .\n\nIt follows from the monotone property of a measure that\n\n\[ \n\mathop{\lim }\limits_{{\epsilon \downarrow 0}}\mu \left( {a + \epsilon, b - \epsilon }\right) = \mu \left( {a, b}\right) = \mu \left...
Yes
Theorem 4.3.4. If every vaguely convergent subsequence of the sequence of s.p.m.’s \( \left\{ {\mu }_{n}\right\} \) converges to the same \( \mu \), then \( {\mu }_{n}\overset{v}{ \rightarrow }\mu \) .
PROOF. To prove the theorem by contraposition, suppose \( {\mu }_{n} \) does not converge vaguely to \( \mu \) . Then by Theorem 4.3.1,(ii), there exists a continuity interval \( \left( {a, b}\right) \) of \( \mu \) such that \( {\mu }_{n}\left( {a, b}\right) \) does not converge to \( \mu \left( {a, b}\right) \) . By ...
Yes
Theorem 4.4.2. Let \( \left\{ {\mu }_{n}\right\} \) and \( \mu \) be p.m.’s. Then \( {\mu }_{n}\overset{v}{ \rightarrow }\mu \) if and only if\n\n\[ \forall f \in {C}_{B} : {\int }_{{\mathcal{R}}^{1}}f\left( x\right) {\mu }_{n}\left( {dx}\right) \rightarrow {\int }_{{\mathcal{R}}^{1}}f\left( x\right) \mu \left( {dx}\ri...
PROOF. Suppose \( {\mu }_{n}\overset{v}{ \rightarrow }\mu \) . Given \( \epsilon > 0 \), there exist \( a \) and \( b \) in \( D \) such that\n\n\[ \text{(7)}\;\mu \left( {(a, b{\rbrack }^{c}}\right) = 1 - \mu (\left( {a, b\rbrack }\right) < \epsilon \text{.} \]\n\nIt follows from vague convergence that there exists \(...
Yes
Theorem 4.4.3. Let a family of p.m.’s \( \left\{ {{\mu }_{\alpha },\alpha \in A}\right\} \) be given on an arbitrary index set \( A \) . In order that every sequence of them contains a subsequence which converges vaguely to a p.m., it is necessary and sufficient that the following condition be satisfied: for any \( \ep...
PROOF. Suppose (11) holds. For any sequence \( \left\{ {\mu }_{n}\right\} \) from the family, there exists a subsequence \( \left\{ {\mu }_{n}^{\prime }\right\} \) such that \( {\mu }_{n}^{\prime }\overset{v}{ \rightarrow }\mu \) . We show that \( \mu \) is a p.m. Let \( J \) be a continuity interval of \( \mu \) which...
Yes
Theorem 4.4.4. If \( \left\{ {\mu }_{n}\right\} \) and \( \mu \) are p.m.’s, then \( {\mu }_{n}\overset{v}{ \rightarrow }\mu \) if and only if one of the two conditions below is satisfied:\n\n(13)\n\n\[ \forall f \in L : \mathop{\lim }\limits_{n}\int f\left( x\right) {\mu }_{n}\left( {dx}\right) \geq \int f\left( x\rig...
PROOF. We begin by observing that the two conditions above are equivalent by putting \( f = - g \) . Now suppose \( {\mu }_{n}\overset{v}{ \rightarrow }\mu \) and let \( {f}_{k} \in {C}_{B},{f}_{k} \uparrow f \) . Then we have\n\n(14)\n\n\[ \mathop{\lim }\limits_{n}\int f\left( x\right) {\mu }_{n}\left( {dx}\right) \ge...
Yes
Theorem 4.4.5. Let \( \\left\\{ {F}_{n}\\right\\}, F \) be the d.f.’s of the r.v.’s \( \\left\\{ {X}_{n}\\right\\}, X \) . If \( {X}_{n} \\rightarrow X \) in pr., then \( {F}_{n}\\overset{v}{ \\rightarrow }F \) . More briefly stated, convergence in pr. implies convergence in dist.
PROOF. If \( {X}_{n} \\rightarrow X \) in pr., then for each \( f \\in {C}_{K} \), we have \( f\\left( {X}_{n}\\right) \\rightarrow f\\left( X\\right) \) in pr. as easily seen from the uniform continuity of \( f \) (actually this is true for any continuous \( f \), see Exercise 10 of Sec. 4.1). Since \( f \) is bounded...
No
Theorem 4.4.6. If \( {X}_{n} \rightarrow X \) in dist, and \( {Y}_{n} \rightarrow 0 \) in dist., then\n\n(a) \( {X}_{n} + {Y}_{n} \rightarrow X \) in dist.\n\n(b) \( {X}_{n}{Y}_{n} \rightarrow 0 \) in dist.
PROOF. We begin with the remark that for any constant \( c,{Y}_{n} \rightarrow c \) in dist. is equivalent to \( {Y}_{n} \rightarrow c \) in pr. (Exercise 4 below). To prove (a), let \( f \in {C}_{K} \) , \( \left| f\right| \leq M \) . Since \( f \) is uniformly continuous, given \( \epsilon > 0 \) there exists \( \del...
No
Theorem 4.5.1. If \( {X}_{n} \rightarrow X \) a.e., then for every \( r > 0 \) :\n\n(1)\n\[ \mathcal{E}\left( {\left| X\right| }^{r}\right) \leq \mathop{\lim }\limits_{{n \rightarrow \infty }}\mathcal{E}\left( {\left| {X}_{n}\right| }^{r}\right) \]
PROOF. (1) is just a case of Fatou's lemma (see Sec. 3.2):\n\n\[ {\int }_{\Omega }{\left| X\right| }^{r}d\mathcal{P} = {\int }_{\Omega }\mathop{\lim }\limits_{n}{\left| {X}_{n}\right| }^{r}d\mathcal{P} \leq \mathop{\lim }\limits_{\frac{}{n}}{\int }_{\Omega }{\left| {X}_{n}\right| }^{r}d\mathcal{P} \]
Yes
Theorem 4.5.2. If \( \left\{ {X}_{n}\right\} \) converges in dist. to \( X \), and for some \( p > 0 \) , \( \mathop{\sup }\limits_{n}{\mathcal{C}}^{c}\left\{ {\left| {X}_{n}\right| }^{p}\right\} = M < \infty \), then for each \( r < p \) : (2) \[ \mathop{\lim }\limits_{{n \rightarrow \infty }}\mathcal{E}\left( {\left|...
PROOF. We prove the second assertion since the first is similar. Let \( {F}_{n}, F \) be the d.f.’s of \( {X}_{n}, X \) ; then \( {F}_{n}\overset{v}{ \rightarrow }F \) . For \( A > 0 \) define \( {f}_{A} \) on \( {\mathcal{R}}^{1} \) as follows: (3) \[ {f}_{A}\left( x\right) = \left\{ \begin{array}{ll} {x}^{r}, & \text...
No
Theorem 4.5.5. Suppose there is a unique d.f. \( F \) with the moments \( \left\{ {{m}^{\left( r\right) }, r \geq }\right. \) \( 1\} \), all finite. Suppose that \( \left\{ {F}_{n}\right\} \) is a sequence of d.f.’s, each of which has all its moments finite:\n\[ {m}_{n}^{\left( r\right) } = {\int }_{-\infty }^{\infty }...
PROOF. Let \( {\mu }_{n} \) be the p.m. corresponding to \( {F}_{n} \) . By Theorem 4.3.3 there exists a subsequence of \( \left\{ {\mu }_{n}\right\} \) that converges vaguely. Let \( \left\{ {\mu }_{{n}_{k}}\right\} \) be any subsequence converging vaguely to some \( \mu \) . We shall show that \( \mu \) is indeed a p...
Yes
Theorem 5.1.1. If the \( {X}_{j} \) ’s are uncorrelated and their second moments have a common bound, then (1) is true in \( {L}^{2} \) and hence also in pr.
This simple theorem is actually due to Chebyshev, who invented his famous inequalities for its proof.
No
Theorem 5.1.2. Under the same hypotheses as in Theorem 5.1.1, (1) holds also a.e.
PROOF. Without loss of generality we may suppose that \( {c}^{e}\left( {X}_{j}\right) = 0 \) for each \( j \), so that the \( {X}_{j} \) ’s are orthogonal. We have by (6):\n\n\[ \mathcal{E}\left( {S}_{n}^{2}\right) \leq {Mn} \]\n\nwhere \( M \) is a bound for the second moments. It follows by Chebyshev’s inequality tha...
Yes
Theorem 5.1.3. Except for a Borel set of measure zero, every number in \( \\left\\lbrack {0,1}\\right\\rbrack \) is simply normal.
PROOF. Consider the probability space \( \\left( {\\mathcal{U},\\mathcal{B}, m}\\right) \) in Example 2 of Sec. 2.2. Let \( Z \) be the subset of the form \( m/{10}^{n} \) for integers \( n \\geq 1, m \\geq 1 \) , then \( m\\left( Z\\right) = 0 \) . If \( \\omega \\in \\mathcal{U} \\smallsetminus Z \), then it has a un...
No
Theorem 5.2.2. Let \( \\left\\{ {X}_{n}\\right\\} \) be pairwise independent and identically distributed r.v.’s with finite mean \( m \) . Then we have\n\n(3)\n\[ \n\\frac{{S}_{n}}{n} \\rightarrow m\\;\\text{ in pr. } \n\]
PROOF. Let the common d.f. be \( F \) so that\n\n\[ \nm = \\mathcal{E}\\left( {X}_{n}\\right) = {\\int }_{-\\infty }^{\\infty }{xdF}\\left( x\\right) ,\\;\\mathcal{E}\\left( \\left| {X}_{n}\\right| \\right) = {\\int }_{-\\infty }^{\\infty }\\left| x\\right| {dF}\\left( x\\right) < \\infty .\n\]\n\nBy Theorem 3.2.1 the ...
No
Theorem 5.2.3. Let \( \left\{ {X}_{n}\right\} \) be a sequence of independent r.v.’s with d.f.’s \( \left\{ {F}_{n}\right\} \) ; and \( {S}_{n} = \mathop{\sum }\limits_{{j = 1}}^{n}{X}_{j} \) . Let \( \left\{ {b}_{n}\right\} \) be a given sequence of real numbers increasing to \( + \infty \) .\n\nSuppose that we have\n...
PROOF OF SUFFICIENCY. Define for each \( n \geq 1 \) and \( 1 \leq j \leq n \) :\n\n\[ \n{Y}_{n, j} = \left\{ \begin{array}{ll} {X}_{j}, & \text{ if }\left| {X}_{j}\right| \leq {b}_{n}; \\ 0, & \text{ if }\left| {X}_{j}\right| > {b}_{n}; \end{array}\right.\n\]\n\nand write\n\n\[ \n{T}_{n} = \mathop{\sum }\limits_{{j = ...
Yes
Theorem 5.3.1. Let \( \\left\\{ {X}_{n}\\right\\} \) be independent r.v.’s such that\n\n\[ \n\\forall n : \\mathcal{E}\\left( {X}_{n}\\right) = 0,\\;\\mathcal{E}\\left( {X}_{n}^{2}\\right) = {\\sigma }^{2}\\left( {X}_{n}\\right) < \\infty .\n\]\n\nThen we have for every \( \\epsilon > 0 \) :\n\n(1)\n\n\[ \n\\mathcal{P}...
PROOF. Fix \( \\epsilon > 0 \) . For any \( \\omega \) in the set\n\n\[ \n\\Lambda = \\left\\{ {\\omega : \\mathop{\\max }\\limits_{{1 \\leq j \\leq n}}\\left| {{S}_{j}\\left( \\omega \\right) }\\right| > \\epsilon }\\right\\}\n\]\n\nlet us define\n\n\[ \nv\\left( \\omega \\right) = \\min \\left\\{ {j : 1 \\leq j \\leq...
Yes
Theorem 5.3.3. Let \( \\left\\{ {X}_{n}\\right\\} \) be independent r.v.’s and define for a fixed constant \( A > 0 \) :\n\n\[ \n{Y}_{n}\\left( \\omega \\right) = \\left\\{ \\begin{array}{ll} {X}_{n}\\left( \\omega \\right) , & \\text{ if }\\left| {{X}_{n}\\left( \\omega \\right) }\\right| \\leq A \\\\ 0, & \\text{ if ...
PROOF. Suppose that the three series converge. Applying Theorem 5.3.1\nto the sequence \( \\left\\{ {{Y}_{n} - \\mathcal{E}\\left( {Y}_{n}\\right) }\\right\\} \), we have for every \( m \\geq 1 \) :\n\n\[ \n\\mathcal{P}\\left\\{ {\\mathop{\\max }\\limits_{{n \\leq k \\leq {n}^{\\prime }}}\\left| {\\mathop{\\sum }\\limi...
Yes
Theorem 5.3.4. If \( \left\{ {X}_{n}\right\} \) is a sequence of independent r.v.’s, then the convergence of the series \( \mathop{\sum }\limits_{n}{X}_{n} \) in pr. is equivalent to its convergence a.e.
PROOF. By Theorem 4.1.2, it is sufficient to prove that convergence of \( \mathop{\sum }\limits_{n}{X}_{n} \) in pr. implies its convergence a.e. Suppose the former; then, given \( \epsilon : 0 < \epsilon < 1 \), there exists \( {m}_{0} \) such that if \( n > m > {m}_{0} \), we have\n\n(8)\n\n\[ \mathcal{P}\left\{ {\le...
Yes
Theorem 5.4.1. Then\n\n\[ \mathop{\sum }\limits_{n}\frac{\mathcal{E}\left( {X}_{n}^{2}\right) }{{a}_{n}^{2}} = \mathop{\sum }\limits_{n}\frac{{\sigma }_{n}^{2}}{{s}_{n}^{2}{\left( \log {s}_{n}\right) }^{1 + {2\epsilon }}} < \infty \]
by Dini's theorem, and consequently\n\n\[ \frac{{S}_{n}}{{s}_{n}{\left( \log {s}_{n}\right) }^{\left( {1/2}\right) + \epsilon }} \rightarrow 0\text{ a.e. } \]
No
Theorem 5.4.2. Let \( \left\{ {X}_{n}\right\} \) be a sequence of independent and identically distributed r.v.'s. Then we have\n\n(8)\n\n\[ \n{\mathcal{E}}^{c}\left( \left| {X}_{1}\right| \right) < \infty \Rightarrow \frac{{S}_{n}}{n} \rightarrow \mathcal{E}\left( {X}_{1}\right) \text{ a.e.,} \n\]
PROOF. To prove (8) define \( \left\{ {Y}_{n}\right\} \) as in (4) with \( {a}_{n} = n \) . Since\n\n\[ \n\mathop{\sum }\limits_{n}\mathcal{P}\left\{ {{X}_{n} \neq {Y}_{n}}\right\} = \mathop{\sum }\limits_{n}\mathcal{P}\left\{ {\left| {X}_{n}\right| > n}\right\} = \mathop{\sum }\limits_{n}\mathcal{P}\left\{ {\left| {X}...
Yes
Theorem 5.5.2. We have\n\n(11)\n\[ \mathop{\lim }\limits_{{t \rightarrow \infty }}\frac{N\left( t\right) }{t} = \frac{1}{m}\text{ a.e. } \]\n\nand\n\[ \mathop{\lim }\limits_{{t \rightarrow \infty }}\frac{\mathcal{E}\{ N\left( t\right) \} }{t} = \frac{1}{m} \]\n\nboth being true even if \( m = + \infty \), provided we t...
PROOF. It follows from (6) that for every \( \omega \) :\n\n\[ \begin{matrix} {S}_{N\left( {t,\omega }\right) }\left( \omega \right) \leq t < {S}_{N\left( {t,\omega }\right) + 1}\left( \omega \right) \end{matrix} \]\n\nand consequently, as soon as \( t \) is large enough to make \( N\left( {t,\omega }\right) > 0 \) ,\n...
Yes
Theorem 5.5.3. Let \( \\left\\{ {{X}_{n}, n \\geq 1}\\right\\} \) be a sequence of independent and identically distributed r.v.’s with finite mean. For \( k \\geq 1 \) let \( {\\mathcal{T}}_{k},1 \\leq k < \\infty \), be the Borel field generated by \( \\left\\{ {{X}_{j},1 \\leq j \\leq k}\\right\\} \). Suppose that \(...
PROOF. Since \( {S}_{0} = 0 \) as usual, we have\n\n(14)\n\[ \n\\text{()}\\;\\mathcal{E}\\left( {S}_{N}\\right) = {\\int }_{\\Omega }{S}_{N}d\\mathcal{P} = \\mathop{\\sum }\\limits_{{k = 1}}^{\\infty }{\\int }_{\\left( N = k\\right) }{S}_{k}d\\mathcal{P} = \\mathop{\\sum }\\limits_{{k = 1}}^{\\infty }\\mathop{\\sum }\\...
Yes
Theorem 6.1.1. Let \( {X}_{1} \) and \( {X}_{2} \) be independent r.v.’s with d.f.’s \( {F}_{1} \) and \( {F}_{2} \) , respectively. Then \( {X}_{1} + {X}_{2} \) has the d.f. \( {F}_{1} * {F}_{2} \) .
PROOF. We wish to show that\n\n(4)\n\n\[ \forall x : \mathcal{P}\left\{ {{X}_{1} + {X}_{2} \leq x}\right\} = \left( {{F}_{1} * {F}_{2}}\right) \left( x\right) . \]\n\nFor this purpose we define a function \( f \) of \( \left( {{x}_{1},{x}_{2}}\right) \) as follows, for fixed \( x \) :\n\n\[ f\left( {{x}_{1},{x}_{2}}\ri...
Yes
Theorem 6.1.2. The convolution of two absolutely continuous d.f.'s with densities \( {p}_{1} \) and \( {p}_{2} \) is absolutely continuous with density \( {p}_{1} * {p}_{2} \) .
PROOF. We have by Fubini's theorem:\n\n\[ \n{\int }_{-\infty }^{x}p\left( u\right) {du} = {\int }_{-\infty }^{x}{du}{\int }_{-\infty }^{\infty }{p}_{1}\left( {u - v}\right) {p}_{2}\left( v\right) {dv} \n\]\n\n\[ \n= {\int }_{-\infty }^{\infty }\left\lbrack {{\int }_{-\infty }^{x}{p}_{1}\left( {u - v}\right) {du}}\right...
Yes
Theorem 6.1.3. For each \( B \in \mathcal{B} \), we have\n\n\[ \left( {{\mu }_{1} * {\mu }_{2}}\right) \left( B\right) = {\int }_{{\mathcal{R}}^{1}}{\mu }_{1}\left( {B - y}\right) {\mu }_{2}\left( {dy}\right) . \]
PROOF. It is easy to verify that the set function \( \left( {{\mu }_{1} * {\mu }_{2}}\right) \left( \cdot \right) \) defined by (7) is a p.m. To show that its d.f. is \( {F}_{1} * {F}_{2} \), we need only verify that its value for \( B = ( - \infty, x\rbrack \) is given by the \( F\left( x\right) \) defined in (3). Thi...
Yes
Corollary. If \( f \) is a ch.f., then so is \( {\left| f\right| }^{2} \) .
To prove the corollary, let \( X \) have the ch.f. \( f \) . Then there exists on some \( \Omega \) (why?) an r.v. \( Y \) independent of \( X \) and having the same d.f., and so also the same ch.f. \( f \) . The ch.f. of \( X - Y \) is\n\n\[ \mathcal{E}\left( {e}^{{it}\left( {X - Y}\right) }\right) = \mathcal{E}\left(...
Yes
Theorem 6.1.5. For each \( \delta > 0 \), we have \( {f}_{\delta } \in {C}_{B}^{\infty } \) . Furthermore if \( f \in {C}_{U} \) , then \( {f}_{\delta } \rightarrow f \) uniformly in \( {\mathcal{R}}^{1} \) .
PROOF. It is easily verified that \( {n}_{\delta } \in {C}_{B}^{\infty } \) . Moreover its \( k \) th derivative \( {n}_{\delta }^{\left( k\right) } \) is dominated by \( {c}_{k,\delta }{n}_{2\delta } \) where \( {c}_{k,\delta } \) is a constant depending only on \( k \) and \( \delta \) so that\n\n\[ \left| {{\int }_{...
Yes
Theorem 6.1.6. If \( \left\{ {\mu }_{n}\right\} \) and \( \mu \) are s.p.m.’s such that\n\n\[ \n\forall f \in {C}_{B}^{\infty } : {\int }_{{\mathcal{R}}^{1}}f\left( x\right) {\mu }_{n}\left( {dx}\right) \rightarrow {\int }_{{\mathcal{R}}^{1}}f\left( x\right) \mu \left( {dx}\right) ,\n\]\n\nthen \( {\mu }_{n}\overset{v}...
This is an immediate consequence of Theorem 4.4.1, and Theorem 6.1.5, if we observe that \( {C}_{0} \subset {C}_{U} \) . The reduction of the class of \
No
Theorem 6.2.1. If \( {x}_{1} < {x}_{2} \), then we have\n\n\[ \mu \left( \left( {{x}_{1},{x}_{2}}\right) \right) + \frac{1}{2}\mu \left( \left\{ {x}_{1}\right\} \right) + \frac{1}{2}\mu \left( \left\{ {x}_{2}\right\} \right) \]\n\n\[ = \mathop{\lim }\limits_{{T \rightarrow \infty }}\frac{1}{2\pi }{\int }_{-T}^{T}\frac{...
PROOF. Observe first that the integrand above is bounded by \( \left| {{x}_{1} - {x}_{2}}\right| \) everywhere and is \( O\left( {\left| t\right| }^{-1}\right) \) as \( \left| t\right| \rightarrow \infty \) ; yet we cannot assert that the \
No
Theorem 6.2.2. If two p.m.'s or d.f.'s have the same ch.f., then they are the same.
PROOF. If neither \( {x}_{1} \) nor \( {x}_{2} \) is an atom of \( \mu \), the inversion formula (4) shows that the value of \( \mu \) on the interval \( \left( {{x}_{1},{x}_{2}}\right) \) is determined by its ch.f. It follows that two p.m.'s having the same ch.f. agree on each interval whose endpoints are not atoms fo...
Yes
Theorem 6.2.3. If \( f \in {L}^{1}\left( {-\infty , + \infty }\right) \), then \( F \) is continuously differentiable, and we have\n\n\[ \n{F}^{\prime }\left( x\right) = \frac{1}{2\pi }{\int }_{-\infty }^{\infty }{e}^{-{ixt}}f\left( t\right) {dt} \n\]
PROOF. Applying (4) for \( {x}_{2} = x \) and \( {x}_{1} = x - h \) with \( h > 0 \) and using \( F \) instead of \( \mu \), we have\n\n\[ \n\frac{F\left( x\right) + F\left( {x - }\right) }{2} - \frac{F\left( {x - h}\right) + F\left( {x - h - }\right) }{2} = \frac{1}{2\pi }{\int }_{-\infty }^{\infty }\frac{{e}^{ith} - ...
Yes
Theorem 6.2.4. For each \( {x}_{0} \), we have\n\n(7)\n\n\[ \mathop{\lim }\limits_{{T \rightarrow \infty }}\frac{1}{2T}{\int }_{-T}^{T}{e}^{-{it}{x}_{0}}f\left( t\right) {dt} = \mu \left( \left\{ {x}_{0}\right\} \right) . \]
PROOF. Proceeding as in the proof of Theorem 6.2.1, we obtain for the integral average on the left side of (7):\n\n(8)\n\n\[ {\int }_{{\mathcal{R}}^{1} - \left\{ {x}_{0}\right\} }\frac{\sin T\left( {x - {x}_{0}}\right) }{T\left( {x - {x}_{0}}\right) }\mu \left( {dx}\right) + {\int }_{\left\{ {x}_{0}\right\} }{1\mu }\le...
Yes
Theorem 6.2.5. We have\n\n\[ \mathop{\lim }\limits_{{T \rightarrow \infty }}\frac{1}{2T}{\int }_{-T}^{T}{\left| f\left( t\right) \right| }^{2}{dt} = \mathop{\sum }\limits_{{x \in {\mathcal{R}}^{1}}}\mu {\left( \{ x\} \right) }^{2}. \]
PROOF. Since the set of atoms is countable, all but a countable number of terms in the sum above vanish, making the sum meaningful with a value bounded by 1 . Formula (9) can be established directly in the manner of (5) and (7), but the following proof is more illuminating. As noted in the proof of the corollary to The...
Yes
Theorem 6.2.6. \( X \) or \( \mu \) is symmetric if and only if its ch.f. is real-valued\n\n(for all \( t \) ).
PROOF. If \( X \) and \( - X \) have the same distribution, they must \
No
Theorem 6.3.1. Let \( \\left\\{ {{\\mu }_{n},1 \\leq n \\leq \\infty }\\right\\} \) be p.m.’s on \( {\\mathcal{R}}^{1} \) with ch.f.’s \( \\left\\{ {{f}_{n},1 \\leq }\\right. \) \( n \\leq \\infty \\} \) . If \( {\\mu }_{n} \) converges vaguely to \( {\\mu }_{\\infty } \), then \( {f}_{n} \) converges to \( {f}_{\\inft...
PROOF. Since \( {e}^{itx} \) is a bounded continuous function on \( {\\mathcal{R}}^{1} \), although complex-valued, Theorem 4.4.2 applies to its real and imaginary parts and yields (1) at once, apart from the asserted uniformity. Now for every \( t \) and \( h \) , we have, as in (ii) of Sec. 6.1:\n\n\[ \n\\left| {{f}_...
Yes
Let \( {\mu }_{n} \) have mass \( \frac{1}{2} \) at 0 and mass \( \frac{1}{2} \) at \( n \) . Then \( {\mu }_{n} \rightarrow {\mu }_{\infty } \), where \( {\mu }_{\infty } \) has mass \( \frac{1}{2} \) at 0 and is not a p.m.
We have\n\n\[ \n{f}_{n}\left( t\right) = \frac{1}{2} + \frac{1}{2}{e}^{\mathrm{{int}}} \n\]\n\nwhich does not converge as \( n \rightarrow \infty \), except when \( t \) is equal to a multiple of \( {2\pi } \).
Yes
Let \( {\mu }_{n} \) be the uniform distribution \( \left\lbrack {-n, n}\right\rbrack \) . Then \( {\mu }_{n} \rightarrow {\mu }_{\infty } \), where \( {\mu }_{\infty } \) is identically zero.
We have\n\n\[ \n{f}_{n}\left( t\right) = \left\{ \begin{matrix} \frac{\sin {nt}}{nt}, & \text{ if }t \neq 0 \\ 1, & \text{ if }t = 0 \end{matrix}\right. \n\]\n\nand\n\n\[ \n{f}_{n}\left( t\right) \rightarrow f\left( t\right) = \left\{ \begin{array}{ll} 0, & \text{ if }t \neq 0 \\ 1, & \text{ if }t = 0 \end{array}\right...
Yes
Theorem 6.3.4. The topologies induced by the two metrics \( \langle {\rangle }_{1} \) and \( \langle {\rangle }_{2} \) on the space of p.m.’s on \( {\mathcal{R}}^{1} \) are equivalent.
This means that for each \( \mu \) and given \( \epsilon > 0 \), there exists \( \delta \left( {\mu ,\epsilon }\right) \) such that: \[ \langle \mu, v{\rangle }_{1} \leq \delta \left( {\mu ,\epsilon }\right) \Rightarrow \langle \mu, v{\rangle }_{2} \leq \epsilon \] \[ \langle \mu, v{\rangle }_{2} \leq \delta \left( {\m...
No
Theorem 6.4.2. If \( F \) has a finite absolute moment of order \( k, k \) an integer \( \geq 1 \), then \( f \) has the following expansion in the neighborhood of \( t = 0 \) :\n\n(3)\n\n\[ f\left( t\right) = \mathop{\sum }\limits_{{j = 0}}^{k}\frac{{i}^{j}}{j!}{m}^{\left( j\right) }{t}^{j} + o\left( {\left| t\right| ...
PROOF. According to a theorem in calculus (see, e.g., Hardy [1], p. 290]), if \( f \) has a finite \( k \) th derivative at the point \( t = 0 \), then the Taylor expansion below is valid:\n\n(4)\n\[ f\left( t\right) = \mathop{\sum }\limits_{{j = 0}}^{k}\frac{{f}^{\left( j\right) }\left( 0\right) }{j!}{t}^{j} + o\left(...
Yes
Theorem 6.4.3. If \( F \) has a finite mean \( m \), then\n\n\[ \frac{{S}_{n}}{n} \rightarrow m\\text{ in pr. } \]
PROOF. Since convergence to the constant \( m \) is equivalent to that in dist. to \( {\\delta }_{m} \) (Exercise 4 of Sec. 4.4), it is sufficient by Theorem 6.3.2 to prove that the ch.f. of \( {S}_{n}/n \) converges to \( {e}^{imt} \) (which is continuous). Now, by (2) of Sec. 6.1 we have\n\n\[ E\\left( {e}^{{it}\\lef...
No
Theorem 6.4.4. If \( F \) has mean \( m \) and finite variance \( {\sigma }^{2} > 0 \), then\n\n\[ \frac{{S}_{n} - {mn}}{\sigma \sqrt{n}} \rightarrow \Phi \text{ in dist. } \]\n\nwhere \( \Phi \) is the normal distribution with mean 0 and variance 1 .
PROOF. We may suppose \( m = 0 \) by considering the r.v.’s \( {X}_{j} - m \), whose\n\nsecond moment is \( {\sigma }^{2} \) . As in the preceding proof, we have\n\n\[ \mathcal{E}\left( {\exp \left( {{it}\frac{{S}_{n}}{\sigma \sqrt{n}}}\right) }\right) = f{\left( \frac{t}{\sigma \sqrt{n}}\right) }^{n} \]\n\n\[ = {\left...
Yes
Theorem 6.4.5. In the notation of Theorem 4.5.5, if (8) there holds together with the following condition:\n\n(6)\n\n\[ \forall t \in {\mathcal{R}}^{1} : \mathop{\lim }\limits_{{k \rightarrow \infty }}\frac{{m}^{\left( k\right) }{t}^{k}}{k!} = 0 \] \nthen \( {F}_{n}\overset{v}{ \rightarrow }F \) .
PROOF. Let \( {f}_{n} \) be the ch.f. of \( {F}_{n} \). For fixed \( t \) and an odd \( k \) we have by the Taylor expansion for \( {e}^{itx} \) with a remainder term:\n\n\[ {f}_{n}\left( t\right) = \int {e}^{itx}d{F}_{n}\left( x\right) = \int \left\{ {\mathop{\sum }\limits_{{j = 0}}^{k}\frac{{\left( itx\right) }^{j}}{...
No
Theorem 6.4.6. Let \( X \) and \( Y \) be independent, identically distributed r.v.’s with mean 0 and variance 1 . If \( X + Y \) and \( X - Y \) are independent then the common distribution of \( X \) and \( Y \) is \( \Phi \) .
PROOF. Let \( f \) be the ch.f., then by \( \left( 1\right) ,{f}^{\prime }\left( 0\right) = 0,{f}^{\prime \prime }\left( 0\right) = - 1 \) . The ch.f. of \( X + Y \) is \( f{\left( t\right) }^{2} \) and that of \( X - Y \) is \( f\left( t\right) f\left( {-t}\right) \) . Since these two r.v.’s are independent, the ch.f....
Yes
Theorem 6.4.7. A ch.f. is that of a lattice distribution if and only if there exists a \( {t}_{0} \neq 0 \) such that \( \left| {f\left( {t}_{0}\right) }\right| = 1 \) .
PROOF. The \
No
Theorem 6.5.1. If \( f \) is positive definite, then for each \( t \in {\mathcal{R}}^{1} \) :\n\n\[ f\left( {-t}\right) = \overline{f\left( t\right) },\;\left| {f\left( t\right) }\right| \leq f\left( 0\right) . \]
PROOF. Taking \( n = 1,{t}_{1} = 0,{z}_{1} = 1 \) in (1), we see that\n\n\[ f\left( 0\right) \geq 0\text{.} \]\n\nTaking \( n = 2,{t}_{1} = 0,{t}_{2} = t,{z}_{1} = {z}_{2} = 1 \), we have\n\n\[ {2f}\left( 0\right) + f\left( t\right) + f\left( {-t}\right) \geq 0 \]\n\nchanging \( {z}_{2} \) to \( i \), we have\n\n\[ f\l...
Yes
Theorem 6.5.3. Let \( f \) on \( {\mathcal{R}}^{1} \) satisfy the following conditions, for each \( t \) :\n\n(9)\n\[ f\left( 0\right) = 1,\;f\left( t\right) \geq 0,\;f\left( t\right) = f\left( {-t}\right) ,\]\n\n\( f \) is decreasing and continuous convex in \( {\mathcal{R}}_{ + } = \lbrack 0,\infty ) \) . Then \( f \...
PROOF. Without loss of generality we may suppose that\n\n\[ f\left( \infty \right) = \mathop{\lim }\limits_{{t \rightarrow \infty }}f\left( t\right) = 0 \]\n\notherwise we consider \( \left\lbrack {f\left( t\right) - f\left( \infty \right) }\right\rbrack /\left\lbrack {f\left( 0\right) - f\left( \infty \right) }\right\...
Yes