Consider i.i.d random variables $X_1,X_2,\ldots,X_n$ having the $U[0,\theta]$ distribution: $$f_{\theta}(x)=\frac{\mathbf1_{[0,\theta]}(x)}{\theta}$$

, where the unknown parameter $\theta\in\{1,2,\ldots\}$.

What can I say about the maximum likelihood estimator (MLE) of $\theta$?

Usually the parameter space is $\mathbb R^{+}$ in which case the MLE is known to be $$X_{(n)}=\max\{X_1,X_2,\ldots,X_n\}$$

, but here the parameter space is restricted to the natural numbers.

In any case, the likelihood function given the sample $x_1,\ldots,x_n$ is

$$L(\theta)=\frac{\mathbf1_{[0,\theta]}(x_1,\ldots,x_n)}{\theta^n}=\frac{\mathbf1_{[x_{(n)},\infty)}(\theta)}{\theta^n}\qquad,\,\theta\in\mathbb N$$

So I should check the values of $L(\theta)$ for each $\theta\in\{1,2,\ldots\}$ and the MLE is that value of $\theta$ for which $L(\theta)$ is maximized. Is this the correct strategy or can I say that MLE of $\theta$ is $\hat\theta=\lfloor X_{(n)}\rfloor$? Or does the MLE exist at all? I am not sure.

Any hint would be helpful.

  • $\begingroup$ $\theta$ must be at least as big as the largest $X_i$ so surely it should equal the next largest integer. $\endgroup$ – user121049 Dec 14 '18 at 17:45
  • $\begingroup$ @user121049 Yes, I should have used 'ceiling' instead of 'floor'. $\endgroup$ – StubbornAtom Dec 14 '18 at 17:49

Given a sample $x\equiv \{x_i\}_{i=1}^n$, the likelihood is $$ L(\theta\mid x)=\theta^{-n}1\{\theta\ge M(x),m(x)\ge 0\}, $$ where $M(x):=\lceil\max_{1\le i\le n}x_i\rceil$ and $m(x):=\min_{1\le i\le n}x_i$. The indicator suggests that $\hat{\theta}_n(x)\ge M(x)$ ($\because$ $L=0$ otherwise). However, taking values larger than $M(x)$ decreases $L$ because of the first term (assuming that $m(x)>0$). Thus, $\hat{\theta}_n(x)= M(x)$.


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.