1
$\begingroup$

Did I mess up the range of integration in my solution to the following problem ?

Consider an experiment for which, conditioned on $\theta,$ the density of $X$ is \begin{align*} f_{\theta}(x) = \frac{2x}{\theta^2},\,\,0 < x< \theta. \end{align*} Suppose the prior density for $\theta$ is \begin{align*} \pi(\theta) = 1,\,\,\,0 \leq \theta \leq 1 \end{align*} Find the posterior density of $\theta,$ then find $\mathbb{E}[\theta|X]$. Do the same for $X = (X_1,\dots, X_n)$ where $X_1,\dots, X_n$ are i.i.d and have the density above.\

The joint density of $\theta$ and $X$ is given by \begin{align*} f_{\theta}(x)\pi(\theta) = \frac{2x}{\theta^2},\,\,0 < x< \theta \leq 1. \end{align*} and so the marginal density $g(x)$ of $X$ is given by \begin{align*} g(x)=\int_{x}^1f_{\theta}(x)\pi(\theta)d\theta &= \int_{x}^1\frac{2x}{\theta^2}d\theta\\ &=2x\int_{x}^1\theta^{-2}d\theta\\ &=2x[-\frac{1}{\theta}]_x^1\\ &= -2(x -1),\,\,\,0 <x<1 \end{align*} So the posterior density of $\theta$ is \begin{align*} f_x(\theta) = \frac{f_{\theta}(x)\pi(\theta)}{g(x)} = \frac{-x}{(x-1)\theta^2}, \,\, x < \theta \leq 1 \end{align*} and \begin{align*} \mathbb{E}[\theta|X]&= \int_{x}^1\frac{-x}{x-1}\theta^{-1}d\theta\\ &=\frac{-x}{x-1}\ln\theta|_x^1\\ &= \frac{x}{x-1}\ln x \end{align*} Now let $X = (X_1,\dots, X_n)$ where each $X_i$ has the density above. Then the joint density is \begin{align*} f_{\theta}(x)\pi(\theta) = \prod_{i = 1}^n\frac{2x_i}{\theta^2},\,\, 0 < x_{[1]} \leq x_{[n]} < \theta \leq 1 \end{align*} and so the marginal density $g(x)$ of $X$ is given by \begin{align*} g(x)=\int_{x_{[n]}}^1f_{\theta}(x)\pi(\theta)d\theta &= \int_{x_{[n]}}^1\prod_{i = 1}^n\frac{2x_i}{\theta^2}d\theta\\ &=\prod_{i = 1}^n2x_i\int_{x_{[n]}}^1\theta^{-2}d\theta\\ &=\prod_{i = 1}^n2x_i[-\frac{1}{\theta}]_{x_{[n]}}^1\\ &=\Bigg(\frac{1}{x_{[n]}} -1\Bigg) \prod_{i = 1}^n2x_i,\,\,\,0 <x<1 \end{align*} and so the posterior density is \begin{align*} f_{x}(\theta) = \Bigg(\prod_{i = 1}^n\frac{2x_i}{\theta^2}\Bigg) \cdot \Bigg( \Bigg(\frac{1}{x_{[n]}} -1\Bigg) \prod_{i = 1}^n2x_i \Bigg)^{-1} \end{align*}

$\endgroup$
1
$\begingroup$

The univariate case seems correct to me. The multivariate case should be as follows: $$\begin{align*} g(x)=\int_{x_{[n]}}^1f_{\theta}(x)\pi(\theta)d\theta &= \int_{x_{[n]}}^1\prod_{i = 1}^n\left(\frac{2x_i}{\theta^2}\right)d\theta\\ &=\left(\prod_{i = 1}^n2x_i\right)\int_{x_{[n]}}^1\theta^{-2n}d\theta\\ &=\left(\prod_{i = 1}^n2x_i\right)\left[-\frac{1}{(2n-1)\theta^{2n-1}}\right]_{x_{[n]}}^1\\ &=\left(\frac{1}{2n-1}\right)\Bigg(\frac{1}{\left(x_{[n]}\right)^{2n-1}} -1\Bigg) \left(\prod_{i = 1}^n2x_i\right),\,\,\,0 <x<1 \end{align*}$$

Then, the posterior is $$\begin{align*} f_{x}(\theta) = \frac{2n-1}{\theta^{2n}} \Bigg(\frac{1}{\left(x_{[n]}\right)^{2n-1}} -1\Bigg)^{-1}, x_{[n]}<\theta\leq 1 \end{align*}$$

$\endgroup$

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.