6
$\begingroup$

$Y_1, Y_2,..., Y_n$ are i.i.d and uniformly distributed on the set $\{1, 2,..., n\}$. Define $X_n = \min\{k: Y_k = Y_j\; \text {for some}\; j < k\}$, and prove that $\frac {X_n}{\sqrt n}$ converges in distribution to a limit with distribution function $F(x) = 1 - \exp\{-(x^2)/2\}$ for $x>0$.

My working:

$$P\left (\frac {X_n}{\sqrt n} \le x\right) = P\left (X_n \le x\sqrt n\right) = 1 - P\left(X_n > x\sqrt n\right)$$ $$= 1 - (1 - 1/n)\cdot(1 - 2/n)\cdot...\cdot(1 - \lfloor {x\sqrt n}\rfloor/n)$$ where $\lfloor\; \rfloor$ is the floor function

so I basically want to show $1 - (1 - 1/n)\cdot(1 - 2/n)\cdot...\cdot(1 - \lfloor {x\sqrt n}\rfloor/n)$ converges to $\exp\{-(x^2)/2\}$ but i have no idea how to do this.

$\endgroup$
3
$\begingroup$

For every $u\leqslant\frac12$, $\mathrm e^{-u-u^2}\leqslant1-u\leqslant\mathrm e^{-u}$ (can you prove this?). Thus, for every $n\geqslant2k$, $$ \exp\left(-\frac{a_k}{n}-\frac{b_k}{n^2}\right)\leqslant\prod_{i=1}^k\left(1-\frac{i}n\right)\leqslant\exp\left(-\frac{a_k}n\right), $$ where $$ a_k=\sum_{i=1}^ki=\frac{k(k+1)}{2},\qquad b_k=\sum_{i=1}^ki^2\leqslant k^3. $$ Let $x\gt0$. If $k=\lfloor x\sqrt{n}\rfloor$, then $n\geqslant2k$ for every $n$ large enough and $k\to\infty$ when $n\to\infty$ hence $a_k\sim\frac12k^2\sim\frac12x^2n$ and $b_k=o( n^2)$, thus the lower and upper bounds of the product both converge to $\mathrm e^{-x^2/2}$.

This shows (and I am rewriting this part of your post because there is a misprint in it) that $$ P[X_n\gt x\sqrt{n}]\to\mathrm e^{-x^2/2},\qquad P[X_n\leqslant x\sqrt{n}]\to1-\mathrm e^{-x^2/2}, $$ hence $X_n/\sqrt{n}\to X$ in distribution, where the density of $X$ is the function $f_X$ defined by $$ f_X(x)=x\mathrm e^{-x^2/2}\mathbf 1_{x\geqslant0}. $$ Alternatively, $X_n^2/(2n)$ converges in distribution to a standard exponential random variable.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.