A random sample of size $n$, $Y_1, Y_2, \ldots, Y_n$ is drawn from a population. The population distribution has the following probability density function

$f(y;\theta)= \begin{cases} \sqrt{\theta}y^{\sqrt{\theta}-1}, & \text{when 0 $\le$ y $\le$ 1} \\ 0, & \text{elsewhere} \end{cases}$

Find an estimator of $\theta$ using the method of moments; call it $\hat \theta$. Prove $\hat \theta$ is a consistent estimator of $\theta$

What I have done so far:

First I opted to find $\mu$. So I set up my integral and solved: $$\mu = \int_0^1 \sqrt{\theta}y^{\sqrt{\theta}}dy = \frac{\sqrt{\theta}}{\sqrt{\theta}+1}$$

Then I solved the equation for $\theta$ to get $\theta = \frac{\mu^2}{(\mu-1)^2}$. So I let the estimator be $\hat{\theta} = \frac{\bar{Y^2}}{(\bar{Y}-1)^2}$

My trouble is coming from how I will prove $\hat{\theta}$ is consistent. I know two ways to show something is consistent, but I am not entirely sure how to apply that in this case. I know if:

1) $\lim \limits_{n \to \infty}E(\hat{\theta})= \theta$ and $\lim \limits_{n \to \infty}V({\hat{\theta}}) = 0$


2) $\lim \limits_{n \to \infty}P(|\hat{\theta_n} - \theta| \le \epsilon)=1$ or $\lim \limits_{n \to \infty}P(|\hat{\theta_n} - \theta| \gt \epsilon)= 0$

Then $\hat{\theta}$ is a consistent estimator for $\theta$. I'm just struggling to make use of those definitions. I'm not sure if my estimator is incorrect, or if I'm forgetting something, but any help would be greatly appreciated.

  • 1
    $\begingroup$ your estimator is correct $\endgroup$
    – LinAlg
    Apr 8, 2018 at 0:59
  • 1
    $\begingroup$ Have you checked the mgf of the mean against known distributions? $\endgroup$
    – LinAlg
    Apr 8, 2018 at 1:06
  • $\begingroup$ @LinAlg We've never been taught beta distributions in class, but after a few minutes of googling, this looks like it might be a beta distribution with $\alpha=\sqrt{\theta}$ and $\beta=1$. Is that correct? This is my first time seeing it, so I may be making some mistakes. $\endgroup$
    – Doadle
    Apr 8, 2018 at 1:26

1 Answer 1


You can combine the (weak) law of large numbers and the continuous mapping theorem. Since $\bar{Y}_n$ converges to $\mu = \dfrac{\sqrt{\theta}}{\sqrt{\theta}+1}$ and the function $g(y) = \dfrac{y^2}{(y-1)^2}$ is continuous on the unit interval $(0,1)$, what can you conclude about $\hat{\theta}_n = g(\bar{Y}_n)$?

Note: The above convergence results are meant to take place in probability, which is what you need to show consistency.


You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .