I'm asked to solve this exercise, but I can't manage to find something satisfying. Any help/hint would be much appreciated.

Let $Y_1, Y_2,\dots, Y_n$ denote a random variable sample of size n from a population whose density is given by :

$$f(y)=\begin{cases}\frac{\alpha y^{\alpha - 1}}{\Theta^\alpha} & 0\leq y \leq \Theta \\ 0 & y \not\in[0,\Theta]\end{cases}$$

Where $\alpha > 0$ is a known fixed value but $\theta$ is unknown.

We consider $\hat{\theta} = \max(Y_1,Y_2,...,Y_n)$. How can one show that $\hat{\theta}$ is a biased estimator for $\theta$.

My try was to compute the distribution function $F(\hat{\theta})$ in order to calculate the probability density function and then simply prove that $E(\hat{\theta}) \neq \theta$ but it didn't get me anywhere.


I suppose the sample is iid. You can find the distribution of $F_{\hat\theta}(t)$ by noticing

$$F_{\hat\theta}(t)=P(\hat\theta\leq t)=P(Y\leq t)^n$$

Thus you can find $f_{\hat\theta}$, and then its expected value

$$E(\hat\theta)=\int_{[0,\Theta]} t f_{\hat\theta}(t)\;\mathrm{d}t$$

Here you have

$$P(Y\leq t)=\int_0^t f_Y(u)\;\mathrm{d}u=\int_0^t \frac{\alpha u^{\alpha-1}}{\Theta^\alpha}\;\mathrm{d}u=\frac{t^\alpha}{\Theta^\alpha}$$




And finally

$$E(\hat\theta)=\int_0^\Theta n\alpha\frac{t^{n\alpha}}{\Theta^{n\alpha}}\;\mathrm{d}t=\frac{n\alpha}{n\alpha+1}\Theta\neq\Theta$$

Hence the estimator $\hat\theta$ is biased.


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.