0
$\begingroup$

I have a sample of size $n$ of independent trials extracted from a Bernoulli distribution with unknown success rate $\theta$. Given this sample, how can I estimate the success rate $\theta$ along with confidence intervals?

$\endgroup$

1 Answer 1

2
$\begingroup$

If we have $n$ iid draws from $Bern(\theta)$ (which is a member of the exponential family) then $\bar X_n$ is complete and sufficient for $\theta$ and since $\mathbb E(\bar X_n) = \theta$ it is the UMVUE. You don't give any details on what your estimator should be like so this seems like a fine initial choice.

Confidence intervals: by the CLT $\sqrt n \frac{\bar X_n - \theta}{\sqrt {\theta(1-\theta)}} \rightarrow_d \mathcal N(0,1)$ so you can appeal to normal theory for a confidence interval if $n$ is sufficiently large and $\theta$ is not too close to 0 or 1. This often doesn't work well in practice so there are other modifications. For instance, you can make use of the fact that $n \bar X_n \sim \ Bin(n, \theta)$. See the Wikipedia article on binomial proportion CIs for more.

$\endgroup$

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.