1
$\begingroup$

Let $X_1, \ldots , X_n$ be i.i.d. sample from the absolute continuous distribution given by the p.d.f. $$f_{\theta}(x) = \begin{cases} \frac{4x^3}{\theta ^4},\qquad \text{if } 0\leq x \leq \theta\\ 0, \qquad \text{ otherwise } \end{cases}\quad,$$ where $\theta > 0$.

Find the maximum likelihood estimator of $\theta$.

So far I got: $$\mathcal{L_\theta}(x)=\prod\limits^n_{i=1}\frac{4x_i^3}{\theta ^4}=\theta^{-4n}4^n \left(\prod\limits^n_{i=1}x_i \right)^3.$$ $$\ln[\mathcal{L_\theta}(x)] \equiv -4n \ln (\theta) + n\ln 4 + 3\sum\limits^n_{i=1}\ln(x_i)$$ \begin{align}{\partial L(\theta) \over \partial \theta} =\frac{-4n}{\theta}=0\end{align} What can be the Maximum likelihood estimator for $\theta$ ? Thanks.

$\endgroup$
  • $\begingroup$ 1. Your likelihood function is incomplete - you've forgotten to include the fact that $\theta \ge x_i$ for each $i$. The complete function should be $\mathcal{L}_\theta \cdot \prod [\theta \ge x_i]$, where $\mathcal{L}_\theta$ is the function above and $[\cdot]$ is the indicator function. 2. In the range $\theta \ge \max x_i$, the corrected function and the original function agree, and you've determined that the likelihood function is positive and strictly decreasing. In the range $\theta < \max x_i$, it is zero once you include the correction. Ergo, the maximiser is $\theta^* = \max x_i$. $\endgroup$ – stochasticboy321 Oct 30 '16 at 21:03
0
$\begingroup$

If $\exists i \in \left\{ 1, \ldots, n\right\}, x_i > \theta, \mathcal{L_θ}(x)=0$

If $\forall i \in \left\{ 1, \ldots, n\right\}, x_i \leq \theta, $

$$\mathcal{L_θ}(x)=\prod\limits^n_{i=1}\frac{4x_i^3}{θ ^4}=θ^{-4n}4^n (\prod\limits^n_{i=1}x_i)^3\geq0.$$

We want $\mathcal{L_θ}(x)$ to be as large as possible and hence we have $\theta \geq \max_i x_i$. Also, When $\theta \geq \max(x_i)$, it is a decreasing function.

Hence $\theta_{ML}=\max(x_i)$

$\endgroup$

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.