Consider a random $n+1$-dimensional vector $v$. Each $v_1,\dots,v_n$ equals $1$ with probability $1/2$ and $-1$ with probability $1/2$ independently. We also set $v_{n+1} = v_1$.

Now consider a random $n$-dimensional vector $w$. Each $w_i$ equals $1$ with probability $1/4$ and $-1$ with probability $1/4$ and equals $0$ with probability $1/2$, all sampled independently.

We know$$P\left(\sum_{i=1}^n v_i w_i = 0\right) = \sum_{k=0}^{\lfloor \frac n2 \rfloor}\frac{n!}{k!(n-2k)!k!}\left(\frac14\right)^k\left(\frac12\right)^{n-2k}\left(\frac14\right)^k \sim \frac{1}{\sqrt{\pi n}}.$$

I wrote computer code to compute probabilities exactly for smallish $n$. It seems numerically that $$P\left(\sum_{i=1}^n v_{i+1} w_i = 0 \; \land \sum_{i=1}^n v_{i} w_i = 0\right) = \frac{\sum _{i=0}^{\lfloor n/2 \rfloor} {2(n-2i) \choose n-2i} {n \choose 2i} {4i \choose 2i}}{ 2^{3n - 1}}.$$

Is this equality true and how could one prove it?

  • $\begingroup$ Could the technique from math.stackexchange.com/questions/1021933/… be applied to your problem maybe? $\endgroup$ – user66307 Jun 4 '15 at 6:52
  • $\begingroup$ @eleanora That would be lovely but I don't see how to do that yet. $\endgroup$ – felipa Jun 6 '15 at 15:31

I havn't found the time to state and simplify a formula, but one would have to do the following. First exclude the possibility that $w_i = 0$. If one or more of the $w_i$'s are Zero, you can just reduce $n$ by that amount. So let's assume $w_i$ are distributed like $v_i$. Now you construct $w_i = u_i v_i$ with $u_i$ being independently sampled like $v_i$. This way $w_i$ will still be Independent of $v_i$ but the two constraints simplify to $\sum_{i=1}^n w_i = 0 \wedge \sum_{i = 1}^n v_i v_{i+1} w_i = 0$. Now the first constraint has only a solution if $n$ is even. In this case there are $\binom{n}{n/2}$ possibilities each of it contains $n/2$ ones and $n/2$ minus ones. Now for fixed $w$ it is possible to calculate the number of Solutions of the equation $\sum_{i = 0}^n x_i w_i$. If $x$ has an even number of $-1$'s it is easy to calculate the probability of $v_i v_{i+1} = x_i$ for all $i$. If $x$ has an odd number of $-1$, the probability is Zero. In this way you might come up with a formula, that could be Close to the one you stated there.


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.