4
$\begingroup$

Let $a_n$ and $b_n$ be two sequences of numbers, both uniformly distributed in $(0,1)$. Then, $$ \lim_{n \to \infty}\frac{1}{n}\sum_{r=1}^{n}f(a_r b_r) = \int_{0}^{1}\int_{0}^{1} f(xy)dxdy $$

Now let $A_n$ and $B_n$ be two increasing sequences of positive integers such that , both the ratios $\frac{A_r}{A_n}$ and $\frac{B_r}{B_n}$, $r = 1,2,\ldots n$ approach uniform distribution in $(0,1)$ as $n \to \infty$. Then,

$$ \lim_{n \to \infty}\frac{1}{n}\sum_{r=1}^{n}f\Big(\frac{A_r B_r}{A_n B_n}\Big) = \int_{0}^{1} f(x^2)dx $$

I found this somewhat unexpected as I would have expected both the above limits to have the same integral form since I was expecting the $a_r b_r$ to have a behavior similar to that of $\frac{A_r}{A_n}\frac{B_r}{B_n}$ but this does not seem to be the case.

Main question: What is the root cause due to which we have two different integral i.e why is the behavior of $a_r b_r$ different from that of $\frac{A_r}{A_n}\frac{B_r}{B_n}$?

Secondary question: Under what conditions or for which functions $f(x) \ne c$, is

$$ \int_{0}^1 \int_{0}^1 f(xy)dxdy = \int_{0}^1 f(x^2)dx $$

$\endgroup$
2
$\begingroup$

The answer to Question 1 has to do with the fact that in the first case, $a_n$ and $b_n$ are unordered, so the joint distribution of $\{(a_r, b_r)\}_{r \ge 1}$ is uniform on the unit square $(0,1)^2$. But in the second expression both $A_r/A_n$ and $B_r/B_n$ are ordered, so their joint distribution is no longer uniform on the unit square. Informally speaking, the product $$\frac{A_r B_r}{A_n B_n}$$ never has a term where $A_r/A_n$ is "small" but $B_r/B_n$ is "large" for a given index $r$, or vice versa. Each is uniform on $(0,1)$ but together, they are highly correlated--in the limit, perfectly correlated.

$\endgroup$
3
$\begingroup$

Take $f(x) = \ln(x)$

Then

$\int_{[0,1]} \int_{[0,1]} f(xy)dxdy = \int_{[0,1]} \int_{[0,1]} \ln(xy) dx dy = \int_{[0,1]} \int_{[0,1]} ( \ln(x) + \ln(y)) dx dy = 2(x\ln(x) - x) |_{0}^{1} = -2 $

And $\int_{[0,1]} f(x^{2})dx = \int_{[0,1]} \ln(x^{2}) dx= \int_{[0,1]} 2\ln(x)dx = 2(x\ln(x) -x)|_{0}^{1} = -2$

$\endgroup$

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.