MLE uniform with varying supports

105 Views Asked by At

We have independent uniform random variables $U_1, U_2, U_3, U_4$ where $U_k ~ [-k\theta,k\theta]$ so these random variables have different supports. Using the observed values, $u_1 = -3.8, u_2 = 8.2, u_3 = -12.6, u_4 = 15.6$, give the maximum likelihood estimate of $\theta$.

I started by determining the likelihood functions for each of the U's. So for U1, we have the support is $[-\theta,\theta]$, and the likelihood function is $$\frac{1}{2\theta^n}$$ Using this same process, the likelihood for U4 is $$\frac{1}{8\theta^n}$$ From these, I can understand that the likelihood function is decreasing, and I have seen in notes that for a uniform random variable with support$(0,\theta)$,$\theta_{MME} = max(X_1,..,X_n)$, but I am having trouble determining how to apply that to this problem.

1

There are 1 best solutions below

2
On

The shortest route and the most rigorous one are the same, which simply require to write correctly each PDF involved, including the support conditions in the PDFs, as they should be.

Here, for every $k$, the PDF $f_k$ of $U_k$ is $$f_k(u)=\frac1{2k\theta}\mathbf 1_{|u|\leqslant k\theta}$$ hence the likelihood of a sample $(U_k)_{1\leqslant k\leqslant n}=(u_k)_{1\leqslant k\leqslant n}$ is $$L(\theta\mid u_1,\ldots,u_n)=\prod_{k=1}^nf_k(u_k)=\prod_{k=1}^n\left(\frac1{2k\theta}\mathbf 1_{|u_k|\leqslant k\theta}\right)=\frac1{2^nn!\theta^n}\mathbf 1_{\theta\geqslant m(u_1,\ldots,u_n)}$$ with $$m(u_1,\ldots,u_n)=\max_{1\leqslant k\leqslant n}\frac{|u_k|}k$$ Thus, $L(\theta\mid u_1,\ldots,u_n)$ is obviously maximal at

$$\hat\theta(u_1,\ldots,u_n)=m(u_1,\ldots,u_n)$$