So I'm working through some computational stats stuff from a free pdf of a book. Specifically I'm looking at their take on the classic Buffon's needle problem. The question has a theoretical part and a computational part. My theoretical background isn't very strong, so I did some research to get those questions answered for me, so I can better understand the computational problem. And the theory seems fairly straight forward on explanation.

My issue is the precursor question to the actual computation problem. The question is this: Let T be the number of crossings in n tosses of the needle, then E=Td/(nl)is an unbiased estimator of 2/π. Calculate the variance of E and thus suggest the best needle length l to use, subject to the restriction l ≤ d.

How would I calculate this? And the best l is simply going to be the one that minimizes the variance of the estimate, right? I understand the definition of variance, but I have no idea how to apply them here.


1 Answer 1


For simplicity I will call your estimator $\theta$

$$\begin{align} \operatorname{var}(\theta) &= \operatorname{var}{[\frac{Td}{nl}}]\\ &= \frac{d^2}{n^2 l^2} \operatorname{var}[T]\\ \end{align}$$

The event of a needle crossing can just be considered a single trial in a bernoulli experiment with $P = \frac{2l}{d \pi}$. Therefore $T$ is binomially distributed with $n=n$, $p = P$.

The variance, $\operatorname{var}[T]$, then just follows: $np*(1-p)$.

Differentiate with respect $l$ then solve.

  • 1
    $\begingroup$ Thank you. This makes perfect sense. I'll have to remind myself to always check for the underlying distribution in question. I thought it was unknown because I was focused on L when in fact I should have been focused on T. $\endgroup$
    – Michael
    Dec 14, 2019 at 17:47

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

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