3
$\begingroup$

I'm currently just learning about Maximum likelihood estimators and I'm a bit stuck on how I should answer this question.

a busy cat

So far I know that for the function to hold, the observation 4 should be <= 2*theta. and that theta>=2 as a result right?

But then I'm not sure how to use it to find the estimator.

I set the Likelihood function as the following

a busy cat

I know I'm supposed to differentiate the log likelihood functions with respect to theta but I get -1/theta=0. Is there no mle estimator or am I missing something?

$\endgroup$
4
$\begingroup$

I'm not sure, but I think it's like finding the MLE for a uniform distribution in that you need to account for the fact that the likelihood of $\theta|\textbf{X}$ equals zero if $2\theta$ is less than the maximum of the sample. Then, I think the likelihood (written with indicator functions rather than piecewise) is written as follows:

$L(\theta|\textbf{x})=[(\frac{3}{4\theta})^{n_1}(\frac{1}{4\theta})^{n-n_1}]I(x_{(1)}\geq0)I(x_{(n)}\leq2\theta)$

where $n_1$ is the number of observations between $0$ and $\theta$.

If that's right, then I don't think you need to differentiate. You just recognize the function is a decreasing function of $\theta$ over $[\frac{x_{(n)}}{2},\infty)$ and equal to zero for values less than $\frac{x_{(n)}}{2}$. The MLE is therefore $\frac{x_{(n)}}{2}$

$\endgroup$
1
  • $\begingroup$ Thanks this really helped me out. This actually helped me out answer another question of mine! $\endgroup$
    – user42668
    May 3 '14 at 13:17
0
$\begingroup$

I don't think the answer is $\hat{\theta} = 2$. However, user44107 is correct: the function here is not everywhere continuous, so it doesn't make sense to differentiate. Try to make its graph for different $\theta$. The OP expression of likelihood is incorrect, BTW. user44107 has the correct expression.

$\endgroup$
1
  • $\begingroup$ Sure it makes sense to differentiate: this log likelihood is differentiable within the interior of the region where it is defined. But, as all Calculus students are admonished--because this is so easily forgotten in the excitement of learning about the power of differentiation--extrema can occur at the boundaries of domains of functions, not just at the critical points, and so the boundary must be tested too. $\endgroup$
    – whuber
    May 2 '14 at 16:40

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.