problem
stringlengths
1
13.6k
solution
stringlengths
0
18.5k
answer
stringlengths
0
575
problem_type
stringclasses
8 values
question_type
stringclasses
4 values
problem_is_valid
stringclasses
1 value
solution_is_valid
stringclasses
1 value
source
stringclasses
8 values
synthetic
bool
1 class
__index_level_0__
int64
0
742k
71. Let for every non-negative random variable $\xi$ the equalities hold $$ \begin{aligned} & L_{*} \xi=\sup \sum_{i} \mathrm{P}\left(A_{i}\right) \inf _{\omega \in A_{i}} \xi(\omega), \\ & L^{*} \xi=\inf \sum_{i} \mathrm{P}\left(A_{i}\right) \sup _{\omega \in A_{i}} \xi(\omega), \end{aligned} $$ where sup and inf ar...
Solution. The inequality $L_{*} \xi \leqslant L^{*} \xi$ follows from $$ \sum_{i} \mathrm{P}\left(A_{i}\right) \inf _{\omega \in A_{i}} \xi(\omega) \leqslant \sum_{i} \mathrm{P}\left(A_{i}\right) \sup _{\omega \in A_{i}} \xi(\omega) $$ (see the definitions of $L_{*} \xi$ and $L^{*} \xi$). Now let the quantity $\xi$ ...
L_{*}\xi=L^{*}\xi=\mathrm{E}\xi
Algebra
proof
Yes
Yes
olympiads
false
33,799
73. (On the connection between Lebesgue and Riemann integration.) Let the Borel function $f=f(x), x \in \mathbb{R}$, be integrable with respect to the Lebesgue measure $\lambda$ $\left(\int_{\mathbb{R}}|f(x)| d \lambda < \infty\right)$. For any $\varepsilon > 0$, there exist (a) a step function $f_{\varepsilon}(x)=\su...
Solution. (a) Without loss of generality, we will assume that $f$ takes non-negative values (in the general case, one should consider the decomposition $f=f^{+}-f^{-}$). Fix any $\varepsilon>0$ and take $n=n_{\varepsilon} \geqslant 1$ such that $$ \int_{f(x) \geqslant n \varepsilon} f(x) d \lambda+\int_{|x|>n} f(x) d ...
proof
Calculus
proof
Yes
Yes
olympiads
false
33,801
7. (See [12].) The Lebesgue integral $\mathrm{E} \xi$ is not defined when $\mathrm{E} \xi^{-} = \mathrm{E} \xi^{+} = \infty$ (see B1.II.6, definition 2). Suppose that for all $a, b > 0$ the limit $$ \overline{\mathrm{E}} \xi = \lim _{n}\left(\mathrm{E}\left[\xi^{+}\right]_{a n} - \mathrm{E}\left[\xi^{-}\right]_{b n}\r...
Solution. (a) Suppose that $\tilde{\mathrm{E}} \xi$ exists and $n \mathrm{P}(|\xi|>n) \rightarrow 0$. For given $a, b>0$, choose (any) integer $c>a \vee b$. Then $$ \begin{aligned} \left|\mathrm{E} \xi I(|\xi|a n)+b n \mathrm{P}(|\xi|>b n)+ \\ +\mathrm{E}|\xi| I(a n\min \{a n, b n\}) \rightarrow 0 \end{aligned} $$ In...
proof
Calculus
proof
Yes
Yes
olympiads
false
33,803
76. Show that the function defined on $[0, \infty)$ $$ f(x)= \begin{cases}1, & x=0 \\ \frac{\sin x}{x}, & x>0\end{cases} $$ is Riemann integrable and the integral of it is the Dirichlet integral $$ (\mathrm{R}) \int_{0}^{\infty} f(x) d x=\frac{\pi}{2} $$ However, establish that the function $f(x)$ is not Lebesgue i...
Solution. Integrating by parts, we find that for $0<a<b$ the equality holds $$ \text { (R) } \int_{a}^{b} \frac{\sin x}{x} d x=\frac{1-\cos b}{b}-\frac{1-\cos a}{a}+(\mathrm{R}) \int_{a}^{b} \frac{1-\cos x}{x^{2}} d x $$ Therefore, for such $a$ and $b$, we have $$ \left|(\mathrm{R}) \int_{a}^{b} \frac{\sin x}{x} d x...
\frac{\pi}{2}
Calculus
proof
Yes
Yes
olympiads
false
33,804
77. Provide an example of a function $f=f(x)$ on $[0,1]$, which is bounded and Lebesgue integrable, but for which it is impossible to find a function $g=g(x)$, Riemann integrable and coinciding with $f=f(x)$ almost everywhere according to the Lebesgue measure on $[0,1]$.
Solution. As $f$, one can take the indicator of a nowhere dense set $\mathscr{C}$ with positive Lebesgue measure. Then the set of discontinuity points of the function $f$ is precisely $\mathscr{C}$, and by the criterion for Riemann integrability (see problem II.6.72), $f$ is not Riemann integrable. On the other hand, t...
proof
Calculus
math-word-problem
Yes
Yes
olympiads
false
33,805
78. Newton used the concept of the integral as an operation inverse to differentiation. In other words, a function $f=f(x)$ on $[a, b]$, $a<b$, was considered integrable if there existed an antiderivative $F=F(x)$ such that $F^{\prime}(x)=f(x)$ on $[a, b]$ (one-sided derivatives are taken at the endpoints $a$ and $b$)....
Solution. As $f(x)$, one should take $F^{\prime}(x)$, where $$ F(x)=x^{2} \sin \frac{1}{x^{2}} $$ Remark. In this context, we will present the construction of the generalized Riemann integral, which includes, as can be shown, the Lebesgue, Riemann, and Newton integrals (see [44]). Specifically, the number $I=(\mathrm...
notfound
Calculus
math-word-problem
Yes
Yes
olympiads
false
33,806
79. When determining the Riemann integral of a function \( f = f(x) \), it is not assumed that this function is Borel measurable. Provide an example of a function \( f(x) \) that is not Borel measurable but is Riemann integrable on the interval \([0,1]\).
Solution. The set of Borel subsets of the interval $[0,1]$ has the power of the continuum $\mathfrak{c}$, while the power of Lebesgue subsets is $2^{\mathfrak{c}}$ (see problem II.2.28). From this, it follows that if $\mathscr{C}$ is a Cantor set on $[0,1]$, then there exists a subset $C \subseteq \mathscr{C}$ which is...
proof
Calculus
math-word-problem
Yes
Yes
olympiads
false
33,807
80. Provide an example of a bounded Borel function $f=f(x, y)$ on $\mathbb{R}^{2}$ such that (for $y \in \mathbb{R}$ and $x \in \mathbb{R}$ respectively) $$ \int_{\mathbb{R}} f(x, y) d x=0, \quad \int_{\mathbb{R}} f(x, y) d y=0 $$ however, this function is not Lebesgue integrable on ( $\mathbb{R}^{2}, \mathscr{B}\lef...
Solution. Take any bounded odd Borel function $g=g(x), x \in \mathbb{R}$, $$ 0<\int_{\mathbb{R}}|g(x)| d x<\infty $$ and define $f(x, y)=g(x-y)$. Then $$ \int_{\mathbb{R}} f(x, y) d x=\int_{\mathbb{R}} g(x-y) d x=\int_{\mathbb{R}} g(x) d x=0 $$ Similarly, it can be established that $\int_{\mathbb{R}} f(x, y) d y=0$...
proof
Calculus
math-word-problem
Yes
Yes
olympiads
false
33,808
81. (On Fubini's Theorem.) Show that Fubini's theorem remains valid if the finiteness of the measures \(\mu_{1}\) and \(\mu_{2}\) involved in its formulation is replaced by their \(\sigma\)-finiteness. The following example shows that without the assumption of \(\sigma\)-finiteness, the statement of Fubini's theorem m...
Solution. The required statement can be derived by a limiting transition from Fubini's theorem for finite measures (which become probability measures after appropriate normalization).
proof
Calculus
proof
Yes
Yes
olympiads
false
33,809
82. Let it be known that for a random variable $\xi$ its expected value $\mathrm{E} \xi$ is negative and $\mathrm{E} e^{\theta \xi}=1$ for some $\theta \neq 0$. Show that then $\theta>0$.
Solution. By Jensen's inequality $e^{\theta \mathrm{E} \xi} \leqslant \mathrm{E} e^{\theta \xi}=1$, which for $\mathrm{E} \xi0$.
proof
Algebra
proof
Yes
Yes
olympiads
false
33,810
83. Let $h=h(t, x)$ be a function defined on the set $[a, b] \times \mathbb{R}$, where $a, b \in \mathbb{R}$ and $a < b$. (a) Suppose that 1) for each $x \in \mathbb{R}$, the function $h(t, x), t \in [a, b]$, is continuous; 2) for each $t \in [a, b]$, the function $h(t, x), x \in \mathbb{R}$, is $\mathscr{B}(\mathbb{...
Solution. (a) The functions $$ \begin{aligned} h_{n}(t, x) & = \\ = & \sum_{i=0}^{n-1} h\left(a+(b-a) \frac{i}{n}, x\right) I_{\left[a+(b-a) \frac{i}{n}, a+(b-a) \frac{i+1}{n}\right]}(t)+h(b, x) I_{\{b\}}(t) \end{aligned} $$ are $\mathscr{B}([a, b]) \times \mathscr{B}(\mathbb{R})$-measurable and converge to $h(t, x)$...
proof
Calculus
proof
Yes
Yes
olympiads
false
33,811
84. (a) Consider the equation $$ Z_{t}=B_{t}+\int_{0}^{t} Z_{s-} d A_{s} $$ where $A_{t}$ and $B_{t}$ are right-continuous (for $t \geqslant 0$) and have left limits (for $t>0$) functions of (locally) bounded variation. Show that in the class of locally bounded functions, this equation has a unique solution $\mathsc...
Solution. (a) Let $C_{t}=A_{t}-\sum_{00$, and $G_{0}=1$ by the formula for integration by parts we get $$ \begin{aligned} \mathscr{E}_{t}(A)=F_{t} G_{t} & =1+\int_{0}^{t} F_{s} d G_{s}+\int_{0}^{t} G_{s-} d F_{s}= \\ & =1+\sum_{00$ for all $0 \leqslant s \leqslant t$ and, secondly, $$ \begin{aligned} Y_{t} \leqslant ...
proof
Calculus
proof
Yes
Yes
olympiads
false
33,812
85. Let $f=f(x)$ be a convex function defined on $\mathbb{R}$. (a) The Legendre transformation of the function $f$ is defined as $$ g(y)=\sup _{x}[x y-f(x)] $$ Establish that $g=g(y)$ is a convex function on $\mathbb{R}$, and the Legendre transformation $\sup _{y}[x y-g(y)]$ of $g$ is nothing but $f=f(x)$. Remark. ...
Solution. (a) The convexity of the function $g$ is established by direct verification. It is easy to deduce from the convexity of the function $f$ that at any point $x \in \mathbb{R}$, $f$ has right $h_{+}(x)$ and left $h_{-}(x)$ derivatives, and $h_{-}(x)$ and $h_{+}(x)$ are non-decreasing in $x$ and $h_{-}(x) \leqsl...
(y)=\int_{0}^{y}F^{-1}(z)=F^{-1}(y)y-\mathrm{E}(F^{-1}(y)-X)^{+}
Algebra
proof
Yes
Yes
olympiads
false
33,813
86. Let $f(x)=\sup _{\alpha}\left(a_{\alpha}^{\top} x+b_{\alpha}\right)$ be a finite function in a convex domain $x \in C \subset \mathbb{R}^{d}$ for some $a_{\alpha} \in \mathbb{R}^{d}, b_{\alpha} \in \mathbb{R}$, where $\alpha$ runs through an arbitrary set of indices. (a) Prove that for any $x, y \in C$, for all $\...
Solution. (a) The statement is verified by elementary checking. (b) We will show that for any convex function $f=f(x)$ and any point $x_{0} \in C$, there exists a hyperplane of the form $y=a^{\top} x+b$, passing through $x_{0}$, such that $$ f(x) \geqslant a^{\top} x+b, \quad x \in C $$ This, of course, will ensure ...
proof
Inequalities
proof
Yes
Yes
olympiads
false
33,814
87. Show that for $a, b \in \mathbb{R}$ and $r \geqslant 0$ the so-called $c_{r}$-inequalities hold $$ |a+b|^{r} \leqslant c_{r}\left(|a|^{r}+|b|^{r}\right) $$ where $c_{r}=1$ for $r<1$ and $c_{r}=2^{r-1}$ for $r \geqslant 1$.
Solution. For $r \geqslant 1$, the given inequality follows from the convexity of the function $f(x)=|x|^{r}, x \in \mathbb{R}$: $$ \left|\frac{a+b}{2}\right|^{r}=f\left(\frac{a+b}{2}\right) \leqslant \frac{f(a)+f(b)}{2}=\frac{|a|^{r}+|b|^{r}}{2} $$ For $0 \leqslant r<1$, the function $g(x)=x^{r}, x \geqslant 0$, is ...
proof
Inequalities
proof
Yes
Yes
olympiads
false
33,815
88. (Weyl's Uncertainty Principle.) Let the random variable $\xi$ have a smooth density $f=f(x)$, and $$ x f(x) \rightarrow 0, \quad|x| \rightarrow \infty $$ Prove that $$ \mathrm{E} \xi^{2} \cdot \mathrm{E}\left|\frac{f^{\prime}(\xi)}{f(\xi)}\right|^{2} \geqslant 1 $$
Solution. The statement follows from the Cauchy-Bunyakovsky inequality: $$ 1=\int_{\mathbb{R}} f(x) d x=-\int_{\mathbb{R}} x f^{\prime}(x) d x=-\mathrm{E} \xi \cdot \frac{f^{\prime}(\xi)}{f(\xi)} \leqslant \sqrt{\mathrm{E} \xi^{2}} \cdot \sqrt{\mathrm{E}\left|\frac{f^{\prime}(\xi)}{f(\xi)}\right|^{2}} $$
proof
Calculus
proof
Yes
Yes
olympiads
false
33,816
90. Let $\xi$ and $\zeta$ be non-negative random variables such that for any $x>0$ the inequality $$ \mathrm{P}(\xi \geqslant x) \leqslant x^{-1} \mathrm{E} \zeta I(\xi \geqslant x) $$ holds. Show that then for any $p>1$ the inequality $$ \mathrm{E} \xi^{p} \leqslant\left(\frac{p}{p-1}\right)^{p} \mathrm{E} \zeta^{p...
Solution. By Fubini's theorem $$ \begin{aligned} \mathrm{E} \xi^{p}=\int_{\mathbb{R}_{+}} p x^{p-1} \mathrm{P}(\xi>x) d x \leqslant \int_{\mathbb{R}_{+}} p x^{p-2} \mathrm{E} \zeta I(\xi \geqslant x) d x= \\ =\mathrm{E} \zeta \int_{0}^{\xi} p x^{p-2} d x=\frac{p \cdot \mathrm{E} \zeta \xi^{p-1}}{p-1} . \end{aligned} $...
proof
Inequalities
proof
Yes
Yes
olympiads
false
33,818
91. Let $X$ and $Y$ be non-negative random variables for which there exist $\alpha > 1$ and $\beta > 0$ such that for all $x > 0$ the inequality $$ \mathrm{P}(X > \alpha x, Y \leqslant x) \leqslant \beta \mathrm{P}(X > x) $$ holds. Let also $f$ be a non-negative increasing function, $f(0) = 0$, and $$ L_{f}(\alpha) ...
Solution. Since $\mathrm{E} f(X)x) d f(x) \geqslant \int_{\mathbb{R}_{+}} \mathbf{P}(X>\alpha x, Y \leqslant x) d f(x)= \\ = & \mathrm{E} \int_{\mathbb{R}_{+}} I(X>\alpha x, Y \leqslant x) d f(x) \geqslant \mathbf{E} \int_{\mathbb{R}_{+}} I(X>\alpha x) d f(x)- \\ & -\mathbb{E} \int_{\mathbb{R}_{+}} I(Y>y) d f(y)=\mathb...
proof
Inequalities
proof
Yes
Yes
olympiads
false
33,819
92. Let $X$ be a random variable with distribution function $F=F(x)$, satisfying the Lipschitz condition: $$ |F(x)-F(y)| \leqslant L|x-y|, \quad x, y \in \mathbb{R}, $$ for some constant $L>0$. Show that $X$ has a density $f=f(x)$ and for almost all $x \in \mathbb{R}$ (with respect to the Lebesgue measure) the inequa...
Solution. Let $\mathrm{P}_{X}$ denote the distribution of the random variable $X$, which is a probability measure on $(\mathbb{R}, \mathscr{B}(\mathbb{R}))$. We will show that this measure is absolutely continuous with respect to the Lebesgue measure $\lambda$ defined on Borel sets. By the regularity of the Lebesgue m...
proof
Calculus
proof
Yes
Yes
olympiads
false
33,820
93. (See [100].) Let the random variable $\xi$ have a unimodal distribution density with a maximum at the point $m=0$, which is non-decreasing to the left and non-increasing to the right of $m$. Such a point of maximum is called the mode or peak of the distribution. (a) Let $g$ be an even function that is increasing o...
Solution. (a) The distribution function $F$ of the quantity $|\xi|$ is concave (since the density $f$ of this quantity decreases on $\mathbb{R}_{+}$), therefore $F(x) \leqslant F(\varepsilon)+f(\varepsilon)(x-\varepsilon)$ for all $x \geqslant 0$. The latter means that $|\xi| \succcurlyeq \zeta \eta$, i.e. $$ \mathrm{...
proof
Inequalities
proof
Yes
Yes
olympiads
false
33,821
94. Let $\xi_{1}, \ldots, \xi_{n}$ be i.i.d. random variables, $\mathrm{P}\left(\xi_{1}>0\right)=1$ and $\mathrm{D} \ln \xi_{1}=\sigma^{2}$. Show that for $\varepsilon>0$ the inequality $$ \mathrm{P}\left(\xi_{1} \cdots \xi_{n} \leqslant\left(\mathrm{E} \xi_{1}\right)^{n} e^{n \varepsilon}\right) \geqslant 1-\frac{\si...
Solution. The statement follows from Chebyshev's inequality $$ \mathrm{P}\left(\left|S_{n}-\mathrm{E} S_{n}\right| \leqslant n \varepsilon\right) \geqslant 1-\frac{\mathrm{D} S_{n}}{n^{2} \varepsilon^{2}} $$ applied to $S_{n}=\sum_{i=1}^{n} \ln \xi_{i}$, as well as Jensen's inequality $\ln \mathrm{E} \xi_{1} \geqslan...
proof
Inequalities
proof
Yes
Yes
olympiads
false
33,822
95. Let $\mathrm{P}$ and $\mathrm{Q}$ be two probability measures on ( $\Omega, \mathscr{F}$ ). Establish the equivalence of the following properties: (1) the measure $Q$ is absolutely continuous with respect to $P$, and for some constant $L \geqslant 1$ the inequality $$ \frac{d \mathbf{Q}}{d \mathrm{P}} \leqslant L...
Solution. In the case when $L=1$ or $\alpha=1$, measures $\mathbf{P}$ and $\mathbf{Q}$ coincide, and thus the equivalence of properties (1) and (2) is obvious. Let us now consider $L>1$, and $\alpha \in(0,1)$. To establish the implication (1) $\Rightarrow$ (2), it is sufficient to take $\alpha=1 / L$ and set $$ \mathb...
proof
Other
proof
Yes
Yes
olympiads
false
33,823
96. Let $\xi$ and $\zeta$ be independent random variables, and $\mathrm{E} \xi=0$. Show that $$ \mathrm{E}|\zeta| \leqslant \mathrm{E}|\xi+\zeta| $$ Generalize this property by establishing that $$ \mathrm{E} \max _{1 \leqslant i \leqslant n}\left|\zeta_{i}\right| \leqslant \mathrm{E} \max _{1 \leqslant i \leqslant ...
Solution. If $\mathrm{E}|\zeta|=\infty$, then $\mathrm{E}|\xi+\zeta| \geqslant \mathrm{E}|\zeta|-\mathrm{E}|\xi|=\infty$ (the finiteness of $\mathrm{E}|\xi|$ is assumed by the condition). Now suppose that $\mathrm{E}|\zeta|<\infty$. Then $\mathrm{E}|\xi+\zeta|$ is also finite. From the independence of $\xi$ and $\eta$,...
proof
Inequalities
proof
Yes
Yes
olympiads
false
33,824
97. Let $X_{1}, X_{2}, \ldots-$ be i.i.d. bounded random variables, $\mathrm{E} X_{1}=0, \mathrm{a}$ $$ S_{n}=X_{1}+\ldots+X_{n}, \quad n \geqslant 1 $$ Prove that for any $p>0$ the following equality holds $$ \mathrm{E}\left|S_{n}\right|^{p}=O\left(n^{p / 2}\right) $$
Solution. Since $$ \sqrt[p]{\mathrm{E}\left|S_{n}\right|^{p}} \leqslant \sqrt[2 m]{\mathrm{ES} S_{n}^{2 m}} $$ for any integer $m \geqslant p$, it is sufficient to prove that $$ E S_{n}^{2 m}=O\left(n^{m}\right) $$ Assume first that $X_{1} \stackrel{d}{=}-X_{1}$. Then $\mathrm{E} X_{i}^{l}=0$ for odd (integer) numb...
proof
Algebra
proof
Yes
Yes
olympiads
false
33,825
98. Let $X_{1}, X_{2}, \ldots$ be i.i.d. random variables. Define $S_{0}=0$, $S_{n}=X_{1}+\ldots+X_{n}$. Define recursively the following ladder indices (also called ladder times): $$ \tau_{0}=0, \quad \tau_{k}=\inf \left\{n>\tau_{k-1}: S_{n}-S_{\tau_{k-1}}>0\right\}, \quad k \geqslant 1, $$ with the usual convention...
Solution. The statement follows from problem II.12.9. Remark. Using ladder indices, one can define the quantity $$ M=\sup \left\{0, S_{1}, S_{2}, \ldots\right\} $$ Specifically, $M=\sup \left\{S_{\tau_{i}}: \tau_{i}0$ we have $$ \mathrm{E} M^{p}=\mathrm{P}\left(\tau_{1}=\infty\right) \sum_{n \geqslant 1} \mathrm{E}...
proof
Algebra
proof
Yes
Yes
olympiads
false
33,826
99. Let (in the notation and conditions of the previous problem) $$ A=\sum_{n \geqslant 1} \frac{\mathrm{P}\left(S_{n} \leqslant 0\right)}{n}, \quad B=\sum_{n \geqslant 1} \frac{\mathrm{P}\left(S_{n}>0\right)}{n} . $$ Show that $$ \mathrm{P}\left(\tau_{1}<\infty\right)= \begin{cases}1, & \text { if } B=\infty \\ 1-e...
$$ \begin{aligned} & \mathrm{P}\left(\tau_{1} \leqslant n\right)=\mathrm{P}\left(\max _{1 \leqslant k \leqslant n} S_{k}>0\right)=\mathrm{P}\left(\max _{1 \leqslant k \leqslant n} S_{k} \geqslant 1\right)=\mathrm{P}\left(\min _{1 \leqslant k \leqslant n} S_{k} \leqslant 0\right) \\ & \mathrm{E} s^{\tau_{1}}=\sum_{n=1}^...
proof
Calculus
proof
Yes
Yes
olympiads
false
33,827
102. Let $\xi_{1}, \xi_{2}, \ldots$ be a sequence of independent random variables uniformly distributed on $[0,1]$. Define $$ \nu=\inf \left\{n \geqslant 2: \xi_{n}>\xi_{n-1}\right\}, \quad \mu(x)=\inf \left\{n \geqslant 1: \xi_{1}+\ldots+\xi_{n}>x\right\}, $$ where $0<n)=x^{n} / n!, n \geqslant 1$ (b) $\nu \stackre...
Solution. (a) For $x \in(0,1]$, the probability $\mathrm{P}(\mu(x)>n)=\mathrm{P}\left(\xi_{1}+\ldots\right.$ $\left.\ldots+\xi_{n} \leqslant x\right)$ is equal to the volume of the $n$-dimensional simplex $\left\{\left(x_{1}, \ldots, x_{n}\right) \in \mathbb{R}_{+}^{n}: x_{1}+\ldots\right.$ $\left.\ldots+x_{n}\xi_{2}>\...
e
Other
proof
Yes
Yes
olympiads
false
33,830
103. Show that for any $\lambda>0$ the Boolean transformation $$ x \rightarrow x-\frac{\lambda}{x} \quad(x \neq 0) $$ preserves the Lebesgue measure on $\mathbb{R}$, i.e., for any integrable function $f$ on $\mathbb{R}$, the following equality holds: $$ \int_{\mathbb{R}} f(x) d x=\int_{\mathbb{R}} f\left(x-\frac{\la...
Solution. Let $\lambda=1$ initially. Then the desired relation follows from the following equalities, valid for any function $f$ integrable on $\mathbb{R}:$ $$ \begin{gathered} \int_{0}^{\infty} f\left(x-\frac{1}{x}\right) d x=\int_{-\infty}^{0} f\left(y-\frac{1}{y}\right) \frac{d y}{y^{2}} \\ \int_{\mathbb{R}} f\left...
proof
Calculus
proof
Yes
Yes
olympiads
false
33,831
104. Let $f=f(x)$ be a convex non-decreasing function, $f(-\infty)=0$. Using the fact that any convex function $f$ has a non-decreasing derivative $f^{\prime}$ almost everywhere, show that $$ f(x)=\int_{\mathbb{R}}(x-u)^{+} d f^{\prime}(u) $$ where $x^{+}=x \vee 0$, and the integral with respect to $d f^{\prime}$ is ...
Solution. For $c>0$ we have $$ \int_{-c}^{\infty}(x-u)^{+} d f^{\prime}(u)=\int_{-c}^{x}(x-u) d f^{\prime}(u)=\int_{-c}^{x} f^{\prime}(u) d u-(x+c) f^{\prime}(-c) $$ From the non-negativity of the left-hand side of the last equality, it follows that for $x>-c$ the inequalities hold $$ 0 \leqslant(x+c) f^{\prime}(-c)...
proof
Calculus
proof
Yes
Yes
olympiads
false
33,832
105. Let $\mathrm{E}|\xi|<\infty$. Show that $$ \mathrm{E} \xi^{2} I(|\xi| \leqslant n)=o(n), \quad n \rightarrow \infty $$
Solution. We have $$ \mathrm{E} \xi^{2} I(|\xi| \leqslant n) \leqslant 2 \int_{0}^{n} x \mathrm{P}(|\xi|>x) d x=o(n), \quad n \rightarrow \infty $$ since $x \mathrm{P}(|\xi|>x) \leqslant \mathrm{E}|\xi| I(|\xi|>x) \rightarrow 0, x \rightarrow \infty$.
proof
Algebra
proof
Yes
Yes
olympiads
false
33,833
106. (See [50].) Let $\xi$ be a random variable. Also, let the function $f=f(x)$ be non-decreasing on $\mathbb{R}$, and the expression $x - f(x) + \mathrm{E} f(\xi) - \mathrm{E} \xi$ changes its sign exactly once from “-” to “+” as $x$ runs from $-\infty$ to $\infty$. Prove that for any continuous convex function $g$ o...
Solution. Let $h(x)=f(x)-\mathrm{E} f(\xi)+\mathrm{E} \xi$. From the convexity of the function $g$ it follows that $$ g(x)-g(h(x)) \geqslant g^{\prime}(h(x))(x-h(x)) $$ where $g^{\prime}$ is the right derivative of the function $g$. By the condition, there exists $x_{0}$ for which $$ x-h(x) \leqslant 0, \quad x<x_{0...
proof
Inequalities
proof
Yes
Yes
olympiads
false
33,834
107. Let $f$ be an arbitrary function for which $$ |f(x)-f(y)| \leqslant|x-y|, \quad x, y \in \mathbb{R} $$ (a) Prove that for any such random variable $\xi$ such that $E \xi^{2}<\infty$, the inequality $$ \mathrm{D} f(\xi) \vee|E \xi f(\xi)-\mathrm{E} \xi \cdot \mathrm{E} f(\xi)| \leqslant \mathrm{D} \xi $$ holds....
Solution. (a) Let $\zeta$ be an independent copy of the random variable $\xi$. Then, as is easy to verify, $$ \mathrm{D} \xi=\mathrm{E} \frac{(\xi-\zeta)^{2}}{2}, \quad \mathrm{E} \xi f(\xi)-\mathrm{E} \xi \cdot \mathrm{E} f(\xi)=\mathrm{E}(\xi-\zeta) \frac{f(\xi)-f(\zeta)}{2} $$ Therefore, $$ \begin{gathered} \math...
proof
Inequalities
proof
Yes
Yes
olympiads
false
33,835
1. Let the random variable $X$ and the $\sigma$-algebra $\mathscr{G}$ be independent (collectively) of the $\sigma$-algebra $\mathscr{E}$, and suppose $\mathrm{E}|X|<\infty$. Show that almost surely $$ \mathrm{E}(X \mid \mathscr{G} \vee \mathscr{E})=\mathrm{E}(X \mid \mathscr{G}) $$ Verify by example that the indepen...
Solution. Since $\mathrm{E}(X \mid \mathscr{G})$ is a $\mathscr{G} \vee \mathscr{E}$-measurable random variable, and the $\sigma$-algebra $\mathscr{G} \vee \mathscr{E}$ is generated by sets of the form $D=BC$, where $B \in \mathscr{G}$ and $C \in \mathscr{E}$, it suffices (for all such $D$) to establish the equality $\...
proof
Algebra
proof
Yes
Yes
olympiads
false
33,836
2. We will say that $\sigma$-algebras $\mathscr{A}$ and $\mathscr{B}$ are conditionally independent relative to $\sigma$-algebra $\mathscr{C}$ if $\mathrm{P}(A B \mid \mathscr{C})=\mathrm{P}(A \mid \mathscr{C}) \mathrm{P}(B \mid \mathscr{C})$ for all $A \in \mathscr{A}$ and $B \in \mathscr{B}$. Show that the condition...
Solution. Obviously, (2) $\Rightarrow(1)$ (it follows to take $X=I_{A}$ ) and (1) $\Rightarrow(3)$. To verify the implication $(3) \Rightarrow(1)$, one needs to use the principle of suitable sets (applying the theorem about $\pi-\lambda$-systems, see theorem 2 in V1.II.2). The equivalence of condition (4) and the condi...
proof
Algebra
proof
Yes
Yes
olympiads
false
33,837
3. Let $\left\{\left(X_{i}, Y_{i}\right)\right\}_{i=1}^{n}$ be independent random vectors (in $\mathbb{R}^{2}$ ). Prove that conditionally on $\mathscr{G}_{n}=\sigma\left(Y_{1}, \ldots, Y_{n}\right)$ the values $X_{1}, \ldots, X_{n}$ are also independent, and $\operatorname{Law}\left(X_{i} \mid \mathscr{G}_{n}\right)=\...
Solution. It is required to establish that for any Borel sets $B_{i}, 1 \leqslant i \leqslant n$, with probability one, the equalities hold: $$ \mathrm{P}\left(\bigcap_{i}\left\{X_{i} \in B_{i}\right\} \mid \mathscr{G}_{n}\right)=\prod_{i} \mathrm{P}\left(X_{i} \in B_{i} \mid \mathscr{G}_{n}\right)=\prod_{i} \mathrm{P...
proof
Other
proof
Yes
Yes
olympiads
false
33,838
4. Let $X^{n}=\left\{X_{i}\right\}_{i=1}^{n}$ be a random sample, i.e., a family of i.i.d. random variables. Suppose that 1) $\tau^{n}=\left(\tau_{1}, \ldots, \tau_{n}\right)$ is a random vector whose components are independent and uniformly distributed on $\{1, \ldots, n\}$; 2) $\sigma^{n}=\left(\sigma_{1}, \ldots, \s...
Solution. Let $f\left(X^{n}, i\right)=X_{i}, 1 \leqslant i \leqslant n$. Then the values $X_{\tau_{i}}=f\left(X^{n}, \tau_{i}\right)$ form a random sample conditionally on $X^{n}$, since due to the independence of $\tau^{n}$ and $X^{n}$ (see the remark to problem II.7.14) with probability one the equality holds $$ \op...
proof
Other
proof
Yes
Yes
olympiads
false
33,839
5. Let $f=f(x)$ be a convex function. Show that for any integrable quantity $\xi$, any $\sigma$-subalgebra $\mathscr{G} \subset \mathscr{F}$, and all $c>0$, the following inequality holds: $$ \mathrm{E} f(\zeta) I(f(\zeta)>c) \leqslant \mathrm{E} 2 f(\xi) I(2 f(\xi)>c) $$ where $\zeta=\mathrm{E}(\xi \mid \mathscr{G})...
Solution. The function $g(x)=(f(x)-c)^{+}$ is convex for any fixed $c>0$, so by Jensen's inequality for conditional expectations (see Problem II.7.10) $$ \mathrm{E}(g(\zeta) \mid \mathscr{G}) \leqslant \mathrm{E}(g(\xi) \mid \mathscr{G}) \quad \text { a.s., } \quad \mathrm{E} g(\zeta) \leqslant \mathrm{E} g(\xi) $$ F...
proof
Inequalities
proof
Yes
Yes
olympiads
false
33,840
7. Let $\xi$ and $\zeta$ be independent random variables. (a) Assuming the existence of the expectation $\mathrm{E} \xi$, show that if the random variables $\xi, \zeta$ are identically distributed, then almost surely (a.s.) $$ \mathrm{E}(\xi \mid \xi+\zeta)=\mathrm{E}(\zeta \mid \xi+\zeta)=\frac{\xi+\zeta}{2} $$ (b)...
Solution. (a) Note that $(\xi, \zeta) \stackrel{d}{=}(\zeta, \xi)$. Therefore, $(\xi, \xi+\zeta) \stackrel{d}{=}$ $\stackrel{d}{=}(\zeta, \xi+\zeta)$ and $\xi \mathbf{1}_{A} \stackrel{d}{=} \zeta \mathbf{1}_{A}$ for any $A \in \sigma(\xi+\zeta)$. Consequently, $\mathrm{E} \xi \mathbf{1}_{A}=\mathrm{E} \zeta \mathbf{1}_...
proof
Calculus
proof
Yes
Yes
olympiads
false
33,842
8. Let $\xi_{1}, \xi_{2}, \ldots$ be i.i.d. random variables, $\mathrm{E}\left|\xi_{1}\right|<\infty$. Show that a.s. $$ \mathrm{E}\left(\xi_{1} \mid S_{n}, S_{n+1}, \ldots\right)=n^{-1} S_{n} $$ where $S_{n}=\xi_{1}+\ldots+\xi_{n}$.
Solution. As in problem II.7.7(a), we should use the fact that $\mathrm{E} \xi_{i} \mathbf{1}_{A}=\mathrm{E} \xi_{i} \mathbf{1}_{A}$ for any $A \in \sigma\left(S_{n}, S_{n+1}, \ldots\right)$ and $i, j \leqslant n$.
proof
Other
proof
Yes
Yes
olympiads
false
33,843
9. (See [92].) Let $\xi$ be a random variable with distribution function $F=F(x)$ such that $F(b)-F(a)>0$ for all $a, b, -\infty<a<b<\infty$. Express $\mathrm{E}(\xi \mid a<\xi \leqslant b)$ in terms of $F$. Establish that $f(a, b)=\mathrm{E}(\xi \mid a<\xi \leqslant b)$ is an increasing function of $a, b$ (on the set...
Solution. According to the definition of conditional mathematical expectation $$ \mathrm{E}(\xi \mid a\lambda \mathrm{E}(\xi \mid a<\xi \leqslant x)+(1-\lambda) \mathrm{E}(\xi \mid a<\xi & \leqslant x)= \\ & =\mathrm{E}(\xi \mid a<\xi \leqslant x) \end{aligned} $$ Similarly, we obtain $\mathrm{E}(\xi \mid a<\xi \leqs...
proof
Calculus
proof
Yes
Yes
olympiads
false
33,844
10. Let $g=g(x)$ be a concave Borel function defined on $\mathbb{R}$ such that $\mathrm{E}|g(\xi)|<\infty$. Show that for conditional expectations a.s. the Jensen's inequality holds: $$ g(\mathrm{E}[\xi \mid \mathscr{G}]) \leqslant \mathrm{E}[g(\xi) \mid \mathscr{G}] $$ and, therefore, $$ g(\mathrm{E} \xi) \leqslant...
Solution. Due to the convexity of the function $g$, there exists a Borel function $l=l(x)$ (as $l(x)$ one can take the left derivative of the function $g$ at the point $x$), such that $$ g(y) \geqslant g(x)+l(x)(y-x), \quad x, y \in \mathbb{R} $$ (for a strictly convex function $g$, the inequality is also strict when...
proof
Inequalities
proof
Yes
Yes
olympiads
false
33,845
11. Let $\xi$ be a random variable, $\mathrm{E}|\xi|<\infty$. Verify that if $\xi \leqslant \mathrm{E}[\xi \mid \mathscr{G}]$ a.s. for some $\sigma$-algebra $\mathscr{G}$, then $\xi=\mathrm{E}[\xi \mid \mathscr{G}]$ a.s.
Solution. Let's take an arbitrary strictly concave increasing function $g$ for which $$ \lim _{x \rightarrow \infty} \frac{g(x)}{x}=\lim _{x \rightarrow-\infty} \frac{g(x)}{x / 2}=1 $$ Then $\mathrm{E} g(\xi) \leqslant \mathrm{E} g(\mathrm{E}[\xi \mid \mathcal{G}])$. Considering Jensen's inequality $\mathrm{E} g(\xi)...
proof
Algebra
proof
Yes
Yes
olympiads
false
33,846
12. Let $\xi$ and $\zeta$ be random variables, $\mathrm{E}|\xi|, \mathrm{E}|\zeta|<\infty, \mathrm{E}[\xi \mid \mathscr{E}] \geqslant \zeta$ and $\mathrm{E}[\zeta \mid \mathscr{G}] \geqslant \xi$ a.s. for some (completed) $\sigma$-algebras $\mathscr{E}$ and $\mathscr{G}$. Show that $\xi$ is $\mathscr{E} \cap \mathscr{G...
Solution. Let the function $g$ be the same as in the solution of problem II.7.11. Then $$ \mathrm{E} g(\mathrm{E}[\xi \mid \mathscr{E}]) \geqslant \mathrm{E} g(\zeta), \quad \mathrm{E} g(\mathrm{E}[\zeta \mid \mathscr{G}]) \geqslant \mathrm{E} g(\xi) $$ Using Jensen's inequality, we get $$ \mathrm{E} g(\mathrm{E}[\x...
proof
Algebra
proof
Yes
Yes
olympiads
false
33,847
13. Let $X, Y$ be random variables, $\mathbf{E}|X|, \mathrm{E}|Y|<\infty$ and $\mathrm{E}(Y \mid X)=0$ a.s. Show that from the condition $\mathrm{E}(Y \mid X+Y)=0$ a.s. it follows that $Y=0$ with probability one.
Solution. Let $f(x)=|x|-\operatorname{arctg}|x|, x \in \mathbb{R}$. The function $f=f(x)$ is increasing on $\mathbb{R}_{+}$ and is an even, strictly convex function on $\mathbb{R}$, and $0 \leqslant f(x) \leqslant|x|$ for all $x \in \mathbb{R}$. By Jensen's inequality for conditional expectations, we have $$ \begin{ga...
0
Algebra
proof
Yes
Yes
olympiads
false
33,848
14. Let the random variable $\xi$ be independent of the $\sigma$-algebra $\mathscr{G}$. Suppose also that $\mathrm{E}|f(\xi, \zeta)|<\infty$ for some Borel function $f=f(x, y)$ and $\mathscr{G}$-measurable random variable $\zeta$. Prove that $\mathrm{E}[f(\xi, \zeta) \mid \mathscr{G}]=g(\zeta)$ a.s., where $g(x)=\mathr...
Solution. It is sufficient to prove that for any $A \in \mathscr{G}$ the conditions $$ \mathrm{E} f(\xi, \zeta) I_{A}=\mathrm{E} g(\zeta) I_{A} \quad \text { and } \quad \mathrm{E} I_{A}|g(\zeta)|<\infty $$ hold. The quantity $\xi$ is independent of ( $\left.\zeta, I_{A}\right)$. Moreover, $\mathrm{E}|f(\xi, \zeta)| ...
proof
Other
proof
Yes
Yes
olympiads
false
33,849
15. Let $X_{1}, X_{2}, \ldots$ be a sequence of independent random variables, $S_{n}=\sum_{i=1}^{n} X_{i}$. Show that $\left(S_{1}, \ldots, S_{n-1}\right)$ and $\left(S_{n+1}, S_{n+2}, \ldots\right)$ are conditionally independent given the $\sigma$-algebra $\sigma\left(S_{n}\right)$.
Solution. Since $$ \xi_{n}=\left(S_{1}, \ldots, S_{n}\right) \quad \text { and } \quad \zeta_{n}=\left(X_{n+1}, X_{n+1}+X_{n+2}, \ldots\right) \text { are independent, } $$ by the remark after problem II.7.14, we have $\mathrm{E}\left[I_{B}\left(\zeta_{n}+S_{n}\right) \mid \xi_{n}\right]=\mathrm{E}\left[I_{B}\left(\...
proof
Algebra
proof
Yes
Yes
olympiads
false
33,850
16. (See [97].) Let $\xi$ be a Bernoulli random variable taking values 0 and 1, and $X$ be some other random variable. Show that the following conditions are equivalent: (1) the random variables $X$ and $\xi$ are conditionally independent given $g(X)$ (where $g=g(x)$ is a Borel function), i.e., with probability one $...
Solution. Let condition (1) be satisfied initially. To prove the validity of condition (2), it is sufficient to establish that $\mathrm{D}(p(X) \mid g(X))=\mathrm{E}\left[p(X)^{2} \mid g(X)\right]-|\mathrm{E}[p(X) \mid g(X)]|^{2}=0$. The last equality follows from the relations $$ \mathrm{E}[\xi \mid g(X)]=\mathrm{E}[...
proof
Other
proof
Yes
Yes
olympiads
false
33,851
17. Show that a random variable $\xi$ and a $\sigma$-algebra $\mathscr{G}$ are independent, i.e., for any $A \in \mathscr{G}$, the variables $\xi$ and $I_{A}$ are independent if and only if $\mathrm{E}[g(\xi) \mid \mathscr{G}]=E g(\xi)$ for every Borel function $g=g(x)$ such that $\mathrm{E}|g(\xi)|<\infty$.
Solution. If $A \in \mathscr{G}$ and $B \in \mathscr{B}(\mathbb{R})$, then from the assumption of independence of $\xi$ and $\mathscr{G}$ we have $\mathrm{P}(A \cap\{g(\xi) \in B\})=\mathrm{P}(A) \mathrm{P}(g(\xi) \in B)$, and thus, $\mathrm{E}[g(\xi) \mid \mathscr{G}]=\mathrm{E} g(\xi)$. Conversely, if this equality h...
proof
Algebra
proof
Yes
Yes
olympiads
false
33,852
18. Let $\xi$ be a non-negative random variable and $\mathscr{G}$ a $\sigma$-subalgebra of the $\sigma$-algebra $\mathscr{F}$. Show that $\mathrm{E}[\xi \mid \mathscr{G}]<\infty$ a.s. if and only if the measure $\mathrm{Q}$, defined on sets $A \in \mathscr{G}$ by $\mathrm{Q}(A)=\int_{A} \xi d \mathrm{P}$, is $\sigma$-f...
Solution. To prove the necessity, let $A_{n}=\{\mathrm{E}[\xi \mid \mathscr{G}] \leqslant n\}$. Then $\mathrm{Q}\left(A_{n}\right)=\int_{A_{n}} \xi d \mathrm{P}=\int_{A_{n}} \mathrm{E}[\xi \mid \mathscr{G}] d \mathrm{P} \leqslant n$, from which we conclude that the measure $\mathrm{Q}$ is $\sigma$-finite, since $\bigcu...
proof
Algebra
proof
Yes
Yes
olympiads
false
33,853
19. Show that conditional probabilities $\mathrm{P}(A \mid B)$ are continuous in the sense that if $\lim _{n} A_{n}=A, \lim _{n} B_{n}=B, \mathrm{P}\left(B_{n}\right)>0, \mathrm{P}(B)>0$, then $\lim _{n} \mathrm{P}\left(A_{n} \mid B_{n}\right)=\mathrm{P}(A \mid B)$
Solution. It is easy to prove that under the given conditions $\lim A_{n} B_{n}=$ $=A B$ (see problem II.1.8). Therefore, according to problem II.1.16, we have $\mathrm{P}\left(A_{n} B_{n}\right) \rightarrow \mathrm{P}(A B), \mathrm{P}\left(B_{n}\right) \rightarrow \mathrm{P}(B)$, from which the required convergence fo...
proof
Calculus
proof
Yes
Yes
olympiads
false
33,854
20. Prove the following version of Fatou's lemma for conditional mathematical expectations. Let $(\Omega, \mathscr{F}, \mathrm{P})$ be a probability space and $\left(\xi_{n}\right)_{n \geqslant 1}$ be a sequence of random variables such that the expectations $\mathrm{E} \xi_{n}, n \geqslant 1$, and $\mathrm{E} \underl...
Solution. Let $\xi_{n}^{k}=\xi_{n}^{+}-\xi_{n}^{-} I\left(\xi_{n}^{-}<k\right), k \geqslant 1$. Then $\xi_{n}^{k} \geqslant \xi_{n}$, $$ \frac{\lim }{n} \mathrm{E}\left(\xi_{n} \mid \mathscr{G}\right) \leqslant \frac{\lim }{n} \mathrm{E}\left(\xi_{n}^{k} \mid \mathscr{G}\right) \quad \text { and } \quad \mathrm{E}\lef...
proof
Calculus
proof
Yes
Yes
olympiads
false
33,855
21. Let, as in Problem II.7.20, $\left(\xi_{n}\right)_{n \geqslant 1}$ be a sequence of random variables for which the expectations $\mathrm{E} \xi_{n}, n \geqslant 1$, are defined, and let $\mathscr{G}$ be a $\sigma$-subalgebra of events from $\mathscr{F}$ such that $$ \varlimsup_{n}^{\lim ^{2}} \mathrm{E}\left(\left...
Solution. Since $\xi_{n}^{+}, \xi_{n}^{-} \leqslant\left|\xi_{n}\right|$ and the function $f(x)=x I(x \geqslant a)$, $a>0$, is monotonic, it follows from condition (*) that (a.s.) $$ \lim _{k \rightarrow \infty} \sup _{n} \mathrm{E}\left(\xi_{n}^{*} I\left(\xi_{n}^{*} \geqslant k\right) \mid \mathscr{G}\right)=0 $$ w...
proof
Calculus
proof
Yes
Yes
olympiads
false
33,856
23. If the family of random variables $\left\{\xi_{n}\right\}_{n \geqslant 1}$ is uniformly integrable and $\xi_{n} \rightarrow \xi$ a.s., then $\mathrm{E} \xi_{n} \rightarrow \mathrm{E} \xi$. At the same time, the a.s. convergence of conditional expectations $\mathrm{E}\left(\xi_{n} \mid \mathscr{G}\right) \rightarrow...
Solution. Let $U, V$ be two independent random variables with uniform distribution on $[0,1]$. Let $\mathscr{G}=\sigma(V)$ and $$ X_{k n}=n I(n U \leqslant 1) I(k-1<n V \leqslant k), \quad 1 \leqslant k \leqslant n $$ Then almost surely (a.s.) $$ \mathrm{E}\left(X_{k n} \mid \mathscr{G}\right)=I(k-1<n V \leqslant k)...
\varlimsup_{n}\mathrm{E}(\xi_{n}\mid\mathscr{G})=1\neq0=\mathrm{E}(\xi\mid\mathscr{G})
Other
proof
Yes
Yes
olympiads
false
33,858
24. Let $\xi_{n} \xrightarrow{L^{p}} \xi$ for some $p \geqslant 1$. Show that then $$ \mathrm{E}\left(\xi_{n} \mid \mathscr{G}\right) \xrightarrow{L^{p}} \mathrm{E}(\xi \mid \mathscr{G}) $$ for any $\sigma$-subalgebra $\mathscr{G} \subseteq \mathscr{F}$.
Solution. By the definition of convergence in $L^{p}$ space and using Jensen's inequality, we obtain $$ \mathrm{E}\left|\mathrm{E}\left(\xi_{n} \mid \mathscr{G}\right)-\mathrm{E}(\xi \mid \mathscr{G})\right|^{p} \leqslant \mathrm{E}\left(\left|\xi_{n}-\xi\right|^{p} \mid \mathscr{G}\right)=\mathrm{E}\left|\xi_{n}-\xi\...
proof
Algebra
proof
Yes
Yes
olympiads
false
33,859
25. Show that $(X, Y) \stackrel{d}{=}(Z, Y)$ if and only if $$ \mathrm{P}(X \in A \mid Y)=\mathrm{P}(Z \in A \mid Y) $$ a.s. for every $A \in \mathscr{B}(\mathbb{R})$.
Solution. Necessity. Let $(X, Y) \stackrel{d}{=}(Z, Y)$. For each $C \in \sigma(Y)$, there exists a Borel set $B \in \mathscr{B}(\mathbb{R})$ such that $C = \{Y \in B\}$ and $$ \begin{aligned} \mathrm{EP}(X \in A \mid Y) I_{C}=\mathrm{E} I(X \in A, Y \in B)=\mathrm{E} I(Z \in A, Y & \in B)= \\ & =\mathrm{EP}(Z \in A \...
proof
Algebra
proof
Yes
Yes
olympiads
false
33,860
26. Provide an example of independent random variables $X$ and $Y$ and a $\sigma$-algebra $\mathscr{G}$ such that on a set of positive measure for some sets $A$ and $B$ the inequality $$ \mathrm{P}(X \in A, Y \in B \mid \mathscr{G}) \neq \mathrm{P}(X \in A \mid \mathscr{G}) \mathrm{P}(Y \in B \mid \mathscr{G}) $$ hol...
Solution. Let $X=\xi+\eta, Y=\xi-\eta$, where $\xi, \eta$ are independent random variables with distribution $\mathscr{N}(0,1)$. Then $X, Y$ are independent Gaussian variables (since $\operatorname{cov}(X, Y)=0$). Define $\mathscr{G}=\sigma(\eta)$. We have $$ \begin{aligned} \mathrm{P}(X \leqslant 0, Y \leqslant 0 \mi...
proof
Other
math-word-problem
Yes
Yes
olympiads
false
33,861
27. Consider the probability space ( $\Omega, \mathscr{F}, \mathrm{P}$ ) as the space $([0,1], \mathscr{B}([0,1]), \lambda)$, where $\lambda$ is the Lebesgue measure. Provide an example of a $\sigma$-subalgebra $\mathscr{G} \subseteq \mathscr{B}([0,1])$ for which the conditional expectation $\mathrm{E}(1 \mid \mathscr{...
Solution. As $\mathscr{G}$, we can take the $\sigma$-subalgebra containing all $B$ from $\mathscr{B}([0,1])$ for which $\lambda(B)=0$ or 1.
proof
Other
math-word-problem
Yes
Yes
olympiads
false
33,862
28. When determining the conditional probability $\mathrm{P}(B \mid \mathscr{G})(\omega)$ of an event $B \in \mathscr{F}$ relative to the $\sigma$-algebra $\mathscr{G} \subseteq \mathscr{F}$, it is usually not assumed that $\mathrm{P}(\cdot \mid \mathscr{G})(\omega)$ is a measure on $(\Omega, \mathscr{F})$ with probabi...
Solution. Let $(\Omega, \mathscr{F}, \mathrm{P})=([0,1], \mathscr{B}([0,1]), \lambda)$, where $\lambda$ is the Lebesgue measure, and $$ \mathscr{G}=\{B \in \mathscr{F}: \lambda(B)=0 \text { or } 1\} . $$ Furthermore, let $\mathrm{P}(B \mid \mathscr{G})(\omega) \equiv \lambda(B)$ when $B \in \mathscr{F}$ is not of the...
proof
Other
proof
Yes
Yes
olympiads
false
33,863
30. Let $\eta$ be a $\mathscr{G}$-measurable random variable, $\xi$ be a $\mathscr{F}$-measurable random variable, and $\mathrm{E}|\eta|^{q}<\infty$, $1 / p + 1 / q = 1$. Show that then $\mathrm{E}(\xi \eta \mid \mathscr{G}) = \eta \mathrm{E}(\xi \mid \mathscr{G})$ a.s.
Solution. Let $\eta^{c}=\eta I(|\eta| \leqslant c), c>0$. Then $\eta^{c}-\mathscr{G}$-measurable and $\mathrm{E}\left(\xi \eta^{c} \mid \mathscr{G}\right)=\eta^{c} \mathrm{E}(\xi \mid \mathscr{G})$ a.s. (by the analogous property for bounded variables). Passing to the limit as $c \rightarrow \infty$, we see that $\eta^...
proof
Algebra
proof
Yes
Yes
olympiads
false
33,865
31. Let $\Omega=\mathbb{Z}_{+}$ and $$ \mathrm{P}_{\lambda}(\omega=k)=e^{-\lambda} \frac{\lambda^{k}}{k!}, \quad k \geqslant 0, $$ be the Poisson distribution on $\Omega$ with parameter $\lambda>0$. Show that for the parameter $1 / \lambda$ there does not exist an unbiased estimator $T=T(\omega)$, i.e., such that for...
Solution. Let $T$ be an unbiased estimator of the parameter $1 / \lambda$. Then $$ e^{\lambda} \lambda \mathrm{E}_{\lambda} T=\sum_{k=1}^{\infty} \frac{T(k-1)}{(k-1)!} \lambda^{k}=e^{\lambda} $$ We have reached a contradiction. Indeed, since $$ \mathrm{E}_{1}|T|=\sum_{k=1}^{\infty} \frac{|T(k-1)|}{(k-1)!}<\infty $$ ...
proof
Algebra
proof
Yes
Yes
olympiads
false
33,866
32. (a) Let $\mathscr{P}=\{\mathrm{P}\}$ be a set of measures on $(\Omega, \mathscr{F})$, i.e., $\mathrm{P} \ll \nu$ for some probability measure $\nu$ and all $\mathrm{P} \in \mathscr{P}$. Show that there exist $\mathrm{P}_{1}, \mathrm{P}_{2}, \ldots \in \mathscr{P}$ such that for any $\mathrm{P} \in \mathscr{P}$ the ...
Solution. (a) As $\mathrm{P}_{n}$, one should take those elements of the class $\mathscr{P}$ for which $$ \operatorname{ess} \sup _{\mathrm{P} \in \mathscr{P}} \frac{d \mathrm{P}}{d \nu}=\sup _{n} \frac{d \mathrm{P}_{n}}{d \nu} $$ up to a set of $\nu$-measure zero (the definition of ess sup is given in problem II.4.2...
proof
Other
proof
Yes
Yes
olympiads
false
33,867
33. Let $(E, \mathscr{E})$ be a Borel space, i.e., a measurable space for which there exists an injective mapping $\varphi: E \rightarrow \mathbb{R}$ such that $\varphi(E) \in \mathscr{B}(\mathbb{R})$ and $\{\varphi(E) \cap B: B \in \mathscr{B}(\mathbb{R})\}=\{\varphi(C): C \in \mathscr{E}\}$. Show that there exists a ...
Solution. Since $\varphi$ is an injection, we have $\varphi^{-1}(\mathscr{G})=\mathscr{E}$, where $$ \mathscr{G}=\{\varphi(E) \cap B: B \in \mathscr{B}(\mathbb{R})\} \subseteq \mathscr{B}(\mathbb{R}) $$ Thus, it suffices to show that $\mathscr{G}=\sigma(\mathscr{A})$ for some countably generated algebra $\mathscr{A}$...
proof
Other
proof
Yes
Yes
olympiads
false
33,868
34. Prove that every Polish space $(S, \rho)$ with metric $\rho$ defines a Borel space $(S, \mathscr{B}(S))$, where $\mathscr{B}(S)$ is the Borel $\sigma$-algebra generated by the open sets. In particular, the spaces $\left(\mathbb{R}^{n}, \mathscr{B}\left(\mathbb{R}^{n}\right)\right), 1 \leqslant n \leqslant \infty$...
Solution. We will divide the proof into several steps.
proof
Other
proof
Yes
Yes
olympiads
false
33,869
35. Let $X$ be a random variable with a symmetric distribution $(X \stackrel{d}{=}-X)$ and the function $\varphi=\varphi(x), x \in \mathbb{R}$, such that $\mathbb{E}[\varphi(X)\|X\|]$ is defined. Show that $$ \mathrm{E}[\varphi(X)|| X \mid]=\frac{1}{2}[\varphi(|X|)+\varphi(-|X|)] \quad \text { a.s. } $$ Using the giv...
Solution. Clearly, we have $(X,|X|) \stackrel{d}{=}(-X,|X|)$, therefore $$ \mathbb{E}[\varphi(X)\|X\|]=\mathbb{E}[\varphi(-X)\|X\| $$ Moreover, $\varphi(|X|)+\varphi(-|X|)=\varphi(X)+\varphi(-X)$. Finally, we obtain $$ 2 \mathbb{E}[\varphi(X) \| X \mid]=\mathbb{E}[\varphi(X)+\varphi(-X) \||X|]=\varphi(|X|)+\varphi(-...
\frac{1}{2}[I(|X|\leqslantx)+1]
Algebra
proof
Yes
Yes
olympiads
false
33,870
36. Let $X$ be a non-negative random variable. Find the conditional probabilities $$ \mathrm{P}(X \leqslant x \mid \lfloor X \rfloor) \quad \text { and } \quad \mathrm{P}(X \leqslant x \mid \lceil X \rceil) $$ where $\lfloor X \rfloor$ is the greatest integer not exceeding $X$ (this number is also denoted by $[X]$), ...
Solution. Let's find the first probability (the second is calculated similarly): $$ \mathrm{P}(X \leqslant x \mid\lfloor X\rfloor)=\varphi(x,\lfloor X\rfloor) $$ where $\varphi(x, n)$ is defined by the conditions $$ \varphi(x, n)= \begin{cases}\frac{\mathrm{P}(X \leqslant x, n \leqslant X<n+1)}{\mathrm{P}(n \leqslan...
notfound
Algebra
math-word-problem
Yes
Yes
olympiads
false
33,871
37. Let $X$ be a random variable with a geometric distribution: $$ \mathrm{P}(X=n)=p q^{n-1}, \quad n \in \mathbb{N}, \quad 0 \leqslant p \leqslant 1, \quad q=1-p $$ Show that for $m, n \in \mathbb{N}$ the following equality holds: $$ \mathrm{P}(X>n+m \mid X>n)=\mathrm{P}(X>m) . $$ Also prove the converse statement...
Solution. The relation $(**)$ is established by direct verification. Let's turn to the second part of the problem. Denoting $f(\cdot)=\mathrm{P}(X>\cdot)$, we find that $f(n+m)=f(n) f(m)$ and $f(n)=f(1)^{n}, n \in \mathbb{N}$. Thus, for natural $n$ with $q=f(1)$ and $p=1-q$, we have $$ \mathrm{P}(X>n+1)-\mathrm{P}(X>n...
proof
Other
proof
Yes
Yes
olympiads
false
33,872
38. (a) If a random variable $X$ is exponentially distributed, then the lack of aftereffect property holds: $$ \mathrm{P}(X>x+y \mid X>x)=\mathrm{P}(X>y), \quad x, y \geqslant 0 $$ Show that if for a non-negative extended (i.e., with values in $[0, \infty]$) random variable $X$ the stated property holds, then one of ...
Solution. (a) Denoting $f(x)=\mathrm{P}(X>x)$, we arrive at the equation $f(x+y)=f(x) f(y)$, which, in particular, guarantees that $f(n x)=f(x)^{n}$ and $f(q)=f(1)^{q}$ for $n \in \mathbb{N}, x \geqslant 0$ and $q \in \mathbb{Q}$. It is not difficult to prove that the specified equation in the class of right-continuous...
proof
Other
proof
Yes
Yes
olympiads
false
33,873
39. Let random variables $X$ and $Y$ have finite second moments. Show that $\operatorname{cov}(X, Y)=\operatorname{cov}(X, \mathrm{E}(Y \mid X))$. Also prove that $\mathrm{D} X \leqslant \mathrm{D} X Y$, when $\mathrm{E}(Y \mid X)=1$.
Solution. Using the fact that $\mathrm{E} X Y=\mathrm{E} X \mathrm{E}(Y \mid X)$ and $\mathrm{E} Y=$ $=\mathrm{E}\mathrm{E}(Y \mid X)$, we obtain the first equality. Moreover, we have $\mathrm{E} X Y=\mathrm{E} X$ and $$ \mathrm{E} X^{2} Y^{2}=\mathrm{E} X^{2} \mathrm{E}\left(Y^{2} \mid X\right) \geqslant \mathrm{E} X...
proof
Algebra
proof
Yes
Yes
olympiads
false
33,874
1. Let $X_{1}, X_{2}, \ldots$ be i.i.d. random variables with the Cauchy distribution with density $$ \frac{1}{\pi\left(1+x^{2}\right)}, \quad x \in \mathbb{R} $$ Show that $$ \frac{M_{n}}{n} \xrightarrow{d} \frac{1}{T} $$ where $M_{n}=\max \left\{X_{1}, \ldots, X_{n}\right\}$, and the random variable $T$ has an ex...
Solution. Note that for all $x \leqslant 0$ the following relation holds $$ \mathrm{P}\left(M_{n} \leqslant n x\right) \leqslant \mathrm{P}\left(M_{n} \leqslant 0\right)=\frac{1}{2^{n}} \rightarrow 0 $$ For $x>0$ we have $$ \mathrm{P}\left(M_{n} \leqslant n x\right)=\left(\frac{\operatorname{arctg} n x}{\pi}+\frac{1...
proof
Other
proof
Yes
Yes
olympiads
false
33,875
2. Let $X_{1}, \ldots, X_{n}, n \geqslant 2,$ be i.i.d. random variables with distribution function $F(x)$ (and density $f(x)$, if it exists) and $M_{n}=\max \left\{X_{1}, \ldots, X_{n}\right\}, m_{n}=\min \left\{X_{1}, \ldots, X_{n}\right\}, R_{n}=M_{n}-m_{n}$. Show that $$ \begin{aligned} F_{m_{n}, M_{n}}(x, y) & =\...
Solution. Using the independence and identical distribution of random variables $X_{1}, \ldots, X_{n}$, we find that $$ \begin{aligned} F_{m_{n}, M_{n}}(x, y) & =\mathrm{P}\left(m_{n} \leqslant x, M_{n} \leqslant y\right)= \\ & =\mathrm{P}\left(M_{n} \leqslant y\right)-\mathrm{P}\left(xx \\ F(y)^{n}, & y \leqslant x\e...
proof
Calculus
proof
Yes
Yes
olympiads
false
33,876
3. Let $X$ and $Y$ be independent Poisson random variables with parameters $\lambda > 0$ and $\mu > 0$ respectively. Show that (a) $X+Y$ has a Poisson distribution with parameter $\lambda + \mu$, (b) the distribution of $X$ conditional on $X+Y$ is binomial: $$ \mathrm{P}(X=k \mid X+Y=n)=C_{n}^{k}\left(\frac{\lambda}...
Solution. Parts (a) and (b) are established by direct verification: \[ \begin{aligned} \mathrm{P}(X+Y=n) & =\sum_{k=0}^{n} \mathrm{P}(X=k, Y=n-k)=\sum_{k=0}^{n} \frac{\lambda^{k} e^{-\lambda}}{k!} \cdot \frac{\mu^{n-k} e^{-\mu}}{(n-k)!}= \\ & =\frac{e^{-(\lambda+\mu)}}{n!} \sum_{k=0}^{n} C_{n}^{k} \lambda^{k} \mu^{n-k...
proof
Algebra
proof
Yes
Yes
olympiads
false
33,877
4. (See [66].) Let $X_{\lambda}$ be a Poisson random variable with parameter $\lambda>0$. (a) Prove the inequalities $$ \begin{array}{ll} \mathrm{P}\left(X_{\lambda} \leqslant n\right) \leqslant \frac{\lambda}{\lambda-n} \cdot \mathrm{P}\left(X_{\lambda}=n\right), & 0 \leqslant n < \lambda \\ \mathrm{P}\left(X_{\lamb...
Solution. (a) We have $$ \begin{aligned} \mathrm{P}\left(X_{\lambda} \geqslant n\right) & =e^{-\lambda} \frac{\lambda^{n}}{n!}\left[1+\frac{\lambda}{n+1}+\frac{\lambda^{2}}{(n+1)(n+2)}+\ldots\right] \leqslant \\ & \leqslant e^{-\lambda} \frac{\lambda^{n}}{n!}\left[1+\frac{\lambda}{n+1}+\frac{\lambda^{2}}{(n+1)^{2}}+\l...
proof
Inequalities
proof
Yes
Yes
olympiads
false
33,878
6. Let $(\Omega, \mathscr{F}, \mathrm{P})$ be a probability space, $\mathscr{A}$ and $\mathscr{B}$ be two $\sigma$-subalgebras in $\mathscr{F}$. Let also $L^{2}(\Omega, \mathscr{G}, \mathrm{P})$ for any $\sigma$-algebra $\mathscr{G}$ be the space of $\mathscr{G}$-measurable random variables with finite second moments. ...
Solution. (a) The equality $\rho^{*}(\sigma(X), \sigma(Y))=\rho^{*}(X, Y)$ follows from the fact that any $\sigma(\xi)$-measurable random variable $\zeta$ can be represented as $\zeta=f(\xi)$ for some Borel function $f$. (b) The relation is established by the same arguments as in Problem II.8.5.
proof
Algebra
proof
Yes
Yes
olympiads
false
33,880
8. Let $X$ and $Y$ be independent random variables uniformly distributed on $[0,1]$. Consider the variable $Z=|X-Y|$. Prove that the distribution $F_{Z}(z)$ has a density $f_{Z}(z)$ and $f_{Z}(z)=2(1-z)$, $z \in[0,1]$. From this, in particular, we can conclude that $E Z=1 / 3$.
Solution. The joint density of the quantities $U$ and $Y$ is $f(u+y) f(y)$, where $U=X-Y$, and $f$ is the density of the quantities $X, Y$. We will compute the density $f_{U}(u)$ for $u \in[0,1]:$ $$ f_{U}(u)=\int_{\mathbb{R}} f(u+y) f(y) d y=\int_{0}^{1-u} d y=1-u $$ The quantity $U$ is symmetric, and $|U|=Z \leqsla...
f_{Z}(z)=2(1-z),\quadz\in[0,1]
Other
proof
Yes
Yes
olympiads
false
33,882
9. (Polar method.) Let the random vector $(U, V)$ have a uniform distribution on the unit circle centered at the origin. Set $$ (X, Y)=(U, V) \sqrt{-\frac{2 \ln \left(U^{2}+V^{2}\right)}{U^{2}+V^{2}}} $$ Show that $X$ and $Y$ are independent and have the distribution $\mathscr{N}(0,1)$.
Solution. Transitioning to polar coordinates $$ (U, V)=\rho(\cos \theta, \sin \theta) $$ we see that the joint density transforms to $$ \pi^{-1} I\left(u^{2}+v^{2} \leqslant 1\right) \longrightarrow 2 r I(r \in[0,1]) \cdot(2 \pi)^{-1} I(\theta \in[0,2 \pi]) $$ i.e., $\theta$ is uniformly distributed on $[0,2 \pi]$,...
proof
Algebra
proof
Yes
Yes
olympiads
false
33,883
11. Let the positive random variable $R$ have a Rayleigh distribution, i.e., for some $\sigma^{2}>0$ the density of $R$ is given by $$ f_{R}(r)=\frac{r}{\sigma^{2}} e^{-\frac{r^{2}}{2 \sigma^{2}}}, \quad r>0 $$ Let also $\theta$ be a random variable independent of $R$ with a uniform distribution on ( $\alpha, \alpha+...
Solution. Note that $$ \cos \theta=\cos 2 \pi\left\{\frac{\theta}{2 \pi}\right\}, \quad \sin \theta=\sin 2 \pi\left\{\frac{\theta}{2 \pi}\right\} $$ and $$ 2 \pi\left\{\frac{\theta}{2 \pi}\right\} \stackrel{d}{=} 2 \pi\left\{\frac{\theta-\alpha}{2 \pi}\right\} \stackrel{d}{=} \varphi $$ where $\varphi$ has a unifor...
proof
Other
proof
Yes
Yes
olympiads
false
33,885
13. Let $F=F(x)$ be a distribution function. Show that for any $a>0$ the functions $$ G(x)=\frac{1}{a} \int_{x}^{x+a} F(u) d u \quad \text { and } \quad H(x)=\frac{1}{2 a} \int_{x-a}^{x+a} F(u) d u $$ are also distribution functions.
Solution. Let the quantity $\xi$ be distributed according to the distribution function $F$, and $\eta$ independently of $\xi$ and has a uniform distribution on $[0,1]$. Then $G$ is the distribution function of the quantity $\xi + a \eta$, and $H$ is the quantity $\xi + 2a \eta - a$ (verified using the convolution formu...
proof
Calculus
proof
Yes
Yes
olympiads
false
33,887
14. Let $X \sim \operatorname{Exp}(\lambda), \lambda>0$. (a) Find the density function of the random variable $Y=X^{1 / \alpha}, \alpha>0$, known as the Weibull distribution. (b) Find the density function of the random variable $Y=\ln X$, known as the double exponential distribution. (c) Show that the integer part $...
Solution. (a) The distribution function of the quantity $Y$ is $$ \mathrm{P}(Y \leqslant y)=\mathrm{P}\left(X \leqslant y^{\alpha}\right)=1-e^{-\lambda y^{\alpha}}, \quad y>0 $$ Differentiating it, we obtain the density function of the quantity $Y$: $$ f(y)=\lambda \alpha y^{\alpha-1} e^{-\lambda y^{\alpha}} I(y \ge...
notfound
Calculus
math-word-problem
Yes
Yes
olympiads
false
33,888
15. (a) Let random variables $X$ and $Y$ have a joint density function $f=f(x, y)$ of the form $f(x, y)=g\left(x^{2}+y^{2}\right)$. Let $R$ and $\theta$ be the polar coordinates: $$ X=R \cos \theta, \quad Y=R \sin \theta $$ Show that $R$ and $\theta$ are independent, and that the variable $\theta$ is uniformly distri...
Solution. (a) By switching to polar coordinates, it is easy to compute the density of the random vector $(R, \theta)$: $$ h(r, \theta)= \begin{cases}r g\left(r^{2}\right), & r \geqslant 0, \\ 0 & \text { otherwise. }\end{cases} $$ Since $h$ can be represented as $h(r, \theta)=p(r) q(\theta)$, the quantities $R$ and $...
proof
Algebra
proof
Yes
Yes
olympiads
false
33,889
16. Let $\left(\xi_{1}, \ldots, \xi_{n}\right)$ be a random vector with density $f=f\left(x_{1}, \ldots\right)$, $$ \mathrm{P}\left(\bigcup_{i<j}\left\{\xi_{i}=\xi_{j}\right\}\right)=\int_{\bigcup_{i<\left\{x_{i}=x_{j}\right\}}} f\left(x_{1}, \ldots, x_{n}\right) d x_{1} \ldots d x_{n}=0 $$ From this, it follows that...
Solution. Let the Borel set $B$ be such that for any $\left(x_{1}, \ldots, x_{n}\right) \in B$ the inequalities $x_{1}<x_{2}<\ldots<x_{n}$ hold. Then $$ \begin{aligned} & \mathrm{P}\left(X_{n} \in B\right)=\sum_{\left(i_{1}, \ldots, i_{n}\right)} \mathrm{P}\left(\left(\xi_{i_{1}}, \ldots, \xi_{i_{n}}\right) \in B\righ...
proof
Other
proof
Yes
Yes
olympiads
false
33,890
17. Let $\xi_{1}, \ldots, \xi_{n}$ be independent and identically distributed (i.i.d.) random variables with a continuous distribution function $F = F(x)$. In this case, $\mathrm{P}\left(\xi_{i}=\xi_{j}\right)=0, i \neq j$ (see problem II.8.57), therefore $\mathrm{P}\left(\xi_{i}=\xi_{j}\right.$ for some $\left.i \neq...
Solution. (a) Due to the continuity of the function $F(x)$, taking into account the independence and identical distribution of the values $\xi_{i}, 1 \leqslant i \leqslant n$, we obtain the formula for $F_{r: n}$: $$ \begin{aligned} F_{r: n}(x) & =\mathrm{P}\left(\xi_{r: n} \leqslant x\right)= \\ & =\mathrm{P}\left(\t...
proof
Other
proof
Yes
Yes
olympiads
false
33,891
18. Let $X_{1}, X_{2}, \ldots$ be a sequence of exchangeable random variables, $$ \mathrm{P}\left(X_{i}=X_{j}\right)=0, \quad i \neq j $$ Let $A_{1}, A_{2}, \ldots$ be a sequence of events such that $A_{1}=\Omega$ and for $n \geqslant 2$ the equality $$ A_{n}=\left\{X_{n}>X_{m} \text { for all } m<n\right\} $$ hold...
Solution. Due to the exchangeability of $X_{n}$, we have $$ \mathrm{P}\left(X_{1}<\ldots<X_{N}\right)=\mathrm{P}\left(B_{\mathrm{i}}\right), \quad B_{\mathrm{i}}=\left\{X_{i_{1}}<\ldots<X_{i_{N}}\right\} $$ for any permutation $\mathbf{i}=\left(i_{1}, \ldots, i_{N}\right)$ of the indices $1, \ldots, N$. Taking into a...
proof
Combinatorics
proof
Yes
Yes
olympiads
false
33,892
19. Let $\xi_{1}, \xi_{2}, \ldots$ be i.i.d. non-negative random variables and $\tau$ be a random variable independent of $\left(\xi_{n}\right)$ and taking values in $\mathbb{N}$. (a) Establish that $\mathrm{E} S_{\tau}=\mathrm{E} \tau \mathrm{E} \xi_{1}$, where $S_{\tau}=\xi_{1}+\ldots+\xi_{\tau}$ and infinite values...
Solution. (a) Due to the independence of $(\xi_{n})$ and $\tau$, we have $$ \mathrm{E} S_{\tau}=\mathrm{E} \sum_{n \geqslant 1} \xi_{n} I(\tau \geqslant n)=\mathrm{E} \xi_{1} \sum_{n \geqslant 1} \mathrm{P}(\tau \geqslant n)=\mathrm{E} \xi_{1} \mathrm{E} \tau $$ (b) First, note that $$ \mathrm{P}\left(S_{\tau}>x\rig...
proof
Other
proof
Yes
Yes
olympiads
false
33,893
20. Let $\xi_{1}, \xi_{2}, \ldots$ be uncorrelated identically distributed random variables with finite second moments, and let $\tau$ be a random variable taking values in $\mathbb{N}$ such that $\mathrm{E} \tau^{2}<\infty$ and $\tau$ is independent of $\xi_{1}, \xi_{2}, \ldots$ Show that $$ \mathrm{E} S_{\tau}=\math...
Solution. From the independence of the sequences $\left\{\xi_{n}\right\}_{n \geqslant 1}$ and $\tau$, it follows that $$ \begin{gathered} \mathrm{E} S_{\tau}^{2} \leqslant \mathrm{E} \tau \sum_{n=1}^{\tau} \xi_{n}^{2}=\mathrm{E} \tau \sum_{n=1}^{\tau} \mathrm{E}\left(\xi_{n}^{2} \mid \tau\right)=\mathrm{E} \tau^{2} \m...
proof
Algebra
proof
Yes
Yes
olympiads
false
33,894
21. (Wiener.) Let $\xi_{1}, \xi_{2}, \ldots$ be i.i.d. random variables, $$ \mathrm{P}\left(\xi_{1} \geqslant 0\right)=1 \quad \text { and } \quad \mathrm{E} \xi_{1} < \infty $$ Define $$ \tau=\inf \left\{n \geqslant 1: \sum_{i=1}^{n} \xi_{i} \geqslant a n\right\} $$ Show that $$ \mathrm{E} \tau < \infty $$ Remar...
Solution. It is easy to see that $$ \lim _{N} \mathrm{E} \frac{1}{N} \sum_{k=0}^{N-1} I_{A_{k, N}}=\lim _{N} \mathrm{P}\left(A_{0, N}\right)=\mathrm{P}\left(\sup _{n}\left[S_{n}-a n\right] \geqslant 0\right) $$ It remains to establish that $$ a \sum_{k=1}^{N} I_{A_{k}} \leqslant \sum_{n=1}^{N} \xi_{n} $$ For any su...
proof
Other
proof
Yes
Yes
olympiads
false
33,895
22. Let $\xi$ and $\eta$ be independent random variables with densities $f_{\xi}=f_{\xi}(x), x \in \mathbb{R}$, and $f_{\eta}(y)=I_{[0,1]}(y)$, i.e., $\eta$ has a uniform distribution on $[0,1]$. Find the densities of the variables $\xi \eta$ and $\xi / \eta$.
Solution. Due to the independence of $\xi$ and $\eta$, we have $$ \begin{aligned} \mathrm{P}(\xi \eta \leqslant z)=\int_{\{u, v: u v \leqslant z\}} f_{\xi}(u) I_{[0,1]}(v) d u d v=\int_{-\infty}^{z} d x & \int_{0}^{1} \frac{f_{\xi}(x / v)}{v} d v= \\ & =\int_{-\infty}^{z} d x \int_{x}^{\infty} \frac{f_{\xi}(y)}{y} d y...
notfound
Calculus
math-word-problem
Yes
Yes
olympiads
false
33,896
2. $\xi \wedge \eta$ is independent of $|\xi-\eta|$, and the density $f$ can be chosen to be positive and continuous on $\mathbb{R}_{+}$. (b) Show that if $\xi$ and $\eta$ are independent exponentially distributed random variables with parameters $\lambda$ and $\mu$, and $\lambda \neq \mu$, then the density of the dis...
Solution. (a) 1. The density of the conditional distribution of $\xi$ given $\xi+\eta=z$ is $$ \left(\int_{0}^{z} f(y) f(z-y) d y\right)^{-1} f(x) f(z-x), \quad 0 \leqslant x \leqslant z $$ and is 0 otherwise. If $\xi$ is exponentially distributed, then the above ratio equals $1 / z$ and, consequently, $\xi$ given $\...
proof
Algebra
proof
Yes
Yes
olympiads
false
33,897
24. Let the random variable $C$ be distributed according to the Cauchy distribution. Show that $$ C \stackrel{d}{=} \frac{X}{Y} \stackrel{d}{=} \frac{X}{|Y|} $$ where $X$ and $Y$ are i.i.d. random variables, distributed according to $\mathscr{N}(0,1)$.
Solution. Due to the symmetry of the quantities $C(C \stackrel{d}{=}-C)$, as well as the quantities $X / Y$ and $X /|Y|$, it is sufficient to prove that $$ |C| \stackrel{d}{=} \frac{|X|}{|Y|} $$ Using the change of variables of the form $x=u y$, for all $z \geqslant 0$ we obtain the equalities $$ \begin{aligned} \ma...
proof
Algebra
proof
Yes
Yes
olympiads
false
33,898
25. Let the random variable $\xi$ take a finite number of values $x_{1}, \ldots, x_{k} \geqslant 0$. Show that $$ \lim _{n \rightarrow \infty}\left(E \xi^{n}\right)^{1 / n}=\max \left\{x_{1}, \ldots, x_{k}\right\} $$
Solution. Let $x=\max \left\{x_{1}, \ldots, x_{k}\right\}$, and $p=\mathrm{P}(\xi=x)>0$. Then $$ x^{n} p \leqslant \mathrm{E} \xi^{n} \leqslant x^{n} \quad \text { and } \quad \lim _{n \rightarrow \infty} p^{1 / n}=1 $$ which leads to the desired equality.
proof
Algebra
proof
Yes
Yes
olympiads
false
33,899
26. Let $\xi$ and $\eta$ be independent random variables taking values in $\mathbb{N}$. Suppose that either $\mathrm{E} \xi < \infty$ or $\mathrm{E} \eta < \infty$. Show that $$ \mathrm{E}(\xi \wedge \eta)=\sum_{n=1}^{\infty} \mathrm{P}(\xi \geqslant n) \mathrm{P}(\eta \geqslant n) $$
Solution. The statement follows from the equality $$ \mathrm{E} \tau=\sum_{n \geqslant 1} n \mathrm{P}(\tau=n)=\sum_{n \geqslant 1} \mathrm{P}(\tau \geqslant n) $$ valid for a quantity $\tau$ with values in $\mathbb{N}$, and also from the fact that $$ \mathrm{P}(\xi \wedge \eta \geqslant n)=\mathrm{P}(\xi \geqslant ...
proof
Algebra
proof
Yes
Yes
olympiads
false
33,900
27. Let $\xi$ and $\eta$ be independent random variables having exponential distributions with parameters $\lambda$ and $\mu$ respectively. Find the distribution functions of the variables $\frac{\xi}{\xi+\eta}$ and $\frac{\xi+\eta}{\xi}$.
Solution. We have $$ \frac{\xi+\eta}{\xi}=\frac{\lambda}{\mu}\left(\frac{\zeta+\eta}{\zeta}\right)-\frac{\lambda}{\mu}+1 $$ where $\zeta=\lambda \xi / \mu \stackrel{d}{=} \eta$. According to problem II.8.23(a), the quantity $(\zeta+\eta)^{-1} \zeta$ is uniformly distributed on the interval $[0,1]$. Therefore, $$ \ma...
\begin{aligned}\mathrm{P}(\frac{\xi+\eta}{\xi}\leqslantx)&=1-(\frac{\mu}{\lambda}(x-1)+1)^{-1},\\\mathrm{P}(\frac{\xi}{\xi+\eta}\leqslantx)&=(\frac{\mu(1-x)}{\lambdax}+1
Algebra
math-word-problem
Yes
Yes
olympiads
false
33,901
28. (See [73].) Let $X$ be a Bernoulli random variable, $$ \mathrm{P}(X=1)=p, \quad \mathrm{P}(X=0)=1-p $$ (a) Show that there exist random variables $Y$ independent of $X$ such that the distribution of their sum is symmetric, i.e., $X+Y \stackrel{d}{=}-(X+Y)$. (b) Establish that the minimum variance $\mathrm{D} Y$ ...
Solution. (a) Let $Y$ be independent of $X$ and $Y \stackrel{d}{=}-X$. Then, obviously, the quantity $X+Y$ is symmetric. (b) Due to the symmetry of the quantity $X+Y$, we have $$ p+\mathrm{E} Y=\mathrm{E}(X+Y)=0 $$ i.e., $\mathrm{E} Y=-p$. Moreover, using the independence of the quantities $X$ and $Y$ and the oddnes...
\mathrm{D}Y\geqslantp(1-p)
Algebra
proof
Yes
Yes
olympiads
false
33,902
29. Let $U$ be a random variable having a uniform distribution on $(0,1)$. Show that (a) for any $\lambda>0$ the variable $-\frac{1}{\lambda} \ln U$ has an exponential distribution with parameter $\lambda$; (b) the variable $\operatorname{tg}\left(\pi U-\frac{\pi}{2}\right)$ has a Cauchy distribution with density $\f...
Solution. (a) Let's compute the distribution function of the quantity $V = -\frac{1}{\lambda} \ln U \geqslant 0$: $$ \mathrm{P}(V \leqslant v) = \mathrm{P}\left(U \geqslant e^{-\lambda v}\right) = \left(1 - e^{-\lambda v}\right) I(v \geqslant 0) $$ Thus, $V$ has an exponential distribution with parameter $\lambda$. ...
proof
Other
proof
Yes
Yes
olympiads
false
33,903
30. Let $\theta$ be a random variable uniformly distributed on $[0,2 \pi)$, and $C$ be a random variable with a Cauchy distribution. Show that $$ \operatorname{ctg} \theta \stackrel{d}{=} \operatorname{ctg} \frac{\theta}{2} \stackrel{d}{=} C \quad \text { and } \quad \cos ^{2} \theta \stackrel{d}{=} \cos ^{2} \frac{\t...
Solution. Due to problem II.8.29, taking into account the symmetry of the quantity $C$, we have $$ C \stackrel{d}{=}-C \stackrel{d}{=}-\operatorname{tg}\left(\pi U-\frac{\pi}{2}\right) \stackrel{d}{=} \operatorname{ctg}(\pi U) \stackrel{d}{=} \operatorname{ctg} \frac{\theta}{2}, $$ where $U$ is uniformly distributed ...
proof
Algebra
proof
Yes
Yes
olympiads
false
33,904
32. For a random variable $\xi$, having a $\mathscr{N}(0,1)$-distribution, $$ \mathrm{P}(\xi \geqslant x) \sim \frac{\varphi(x)}{x}, \quad x \rightarrow \infty, \quad \text { where } \varphi(x)=\frac{1}{\sqrt{2 \pi}} e^{-\frac{x^{2}}{2}} . $$ Find the corresponding asymptotic for a random variable $\zeta$, having a g...
Solution. The desired asymptotic is $$ \mathrm{P}(\zeta \geqslant x)=\frac{1}{\Gamma(\alpha)} \int_{x}^{\infty} z^{\alpha-1} e^{-z} d z \sim f(x)=\frac{x^{\alpha-1} e^{-x}}{\Gamma(\alpha)}, \quad x \rightarrow \infty $$ Indeed, by L'Hôpital's rule $$ \begin{aligned} \lim _{x \rightarrow \infty} \frac{\int_{x}^{\inft...
\frac{x^{\alpha-1}e^{-x}}{\Gamma(\alpha)},\quadxarrow\infty
Calculus
math-word-problem
Yes
Yes
olympiads
false
33,906
33. Let $X$ and $Y$ be two non-degenerate independent random variables such that their product $X Y$ has a discrete distribution. Show that each of the variables $X$ and $Y$ has a discrete distribution.
Solution. Let us show, for example, that $X$ has a discrete distribution. Represent the distribution function $F=F(x)$ of the quantity $X$ in the form $F(x)=G(x)+H(x)$, where $x \in \mathbb{R}$, $$ G(x)=\sum_{u \leqslant x}[F(u)-F(u-)] $$ - the jump function of the function $F$, and $H(x)=F(x)-G(x)$ - the continuous ...
proof
Algebra
proof
Yes
Yes
olympiads
false
33,907
34. Let $I$ be an open set in $\mathbb{R}^{n}$ and $y=\varphi(x)$ be a function defined on $I$, with values in $\mathbb{R}^{n}$. (If $x=\left(x_{1}, \ldots, x_{n}\right) \in I$, then $y=\left(y_{1}, \ldots, y_{n}\right)$, where $y_{i}=\varphi_{i}\left(x_{1}, \ldots, x_{n}\right), i=1, \ldots, n$.) It is assumed that al...
Solution. To establish the desired formula, one should apply the multidimensional analogue of the integration by substitution formula (see the remark to Problem II.6.15).
proof
Calculus
proof
Yes
Yes
olympiads
false
33,908
35. Let $Y=A X+b$, where $X=\left(X_{1}, \ldots, X_{n}\right), Y=\left(Y_{1}, \ldots, Y_{n}\right)$, matrix $A$ of order $n \times n$ is such that $|\operatorname{det} A|>0$, and $b$ is an $n$-dimensional vector. Show that $$ f_{Y}(y)=\frac{1}{|\operatorname{det} A|} f_{X}\left(A^{-1}(y-b)\right) $$
Solution. The statement follows from the result of Problem II.8.34 if we take $\varphi(x)=A x+b$ and note that $\left|J_{\varphi^{-1}}(y)\right|=1 /|\operatorname{det} A|$.
proof
Algebra
proof
Yes
Yes
olympiads
false
33,909
37. (a) Show that the following functions: $$ \begin{aligned} & F_{\mathrm{G}}(x)=\exp \left(-e^{-x}\right), \quad x \in \mathbb{R} ; \\ & F_{\mathrm{F}}(x, \alpha)=\left\{\begin{array}{ll} 0, & x \leq 0 ; \\ \exp \left(-x^{-\alpha}\right), & x > 0 ;\end{array}\right. \\ & F_{\mathrm{W}}(x, \alpha)=\left\{\begin{array...
Solution. Point (a) is obvious. Let's establish point (b): $$ \begin{aligned} \mathrm{P}(\ln X \leqslant x) & =F_{\mathrm{F}}\left(a e^{x}, \alpha\right)=F_{\mathrm{G}}(\alpha x+\alpha \ln a) \\ \mathrm{P}(-\ln (-Y) \leqslant x) & =F_{\mathrm{W}}\left(-a e^{-x}, \alpha\right)=F_{\mathrm{G}}(\alpha x-\alpha \ln a) \end...
proof
Algebra
proof
Yes
Yes
olympiads
false
33,911
38. Let $U, V$ be independent random variables, where $U$ has a uniform distribution on the interval $[0,1]$, and $\{u\}$ is the fractional part of the number $u$. Show that the random variable $\{U+V\}$ is uniformly distributed on $[0,1]$ and is independent of $V$.
Solution. First, let's note the following easily verifiable fact. If a random variable $Z$ is uniformly distributed on $[z, 1+z]$, where $z \in \mathbb{R}$, then its fractional part $\{Z\}$ is uniformly distributed on $[0,1]$. Due to the independence of $U$ and $V$, the variable $U+V$ conditioned on $V=v$ is uniforml...
proof
Algebra
proof
Yes
Yes
olympiads
false
33,912