I would like to show that the expression $ \frac{E\,\left[\, p^{t+1}\left(1-p\right)^{\left(n-t\right)}\right]}{E\left[\, p^{t}\left(1-p\right)^{\left(n-t\right)}\right]} $ , where $p$ is random on $[0, 1] $ and $n$ and $t \leq n$ are positive integers, is increasing in $t$ (for any $n$).

The background of this question is as follows. I'm currently starting to learn Bayesian statistics and I'm trying to understand some simple models. The above expression arises when there is a Bernoulli experiment with unknown probability $p$. Suppose that after $n$ trials we have drawn $t$ successes. It seems very intuitive that the larger $t$, the higher the posterior expectation of $p$ should be. However, the posterior expectation can be expressed as the expression above and I do not find a straightforward way to show that this expression is increasing in $t$.

I assume that this is a simple textbook problem, but all I can find are treatments in which a specific prior distribution of $p$ is assumed. I however would like to see a solution which allows $p$ to have any distribution on $[0, 1]$, as long as the above expectations are defined.


Your Answer

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

Browse other questions tagged or ask your own question.