1
$\begingroup$

Suppose $\hat{\theta}_n$ is the MLE for some parameter $\theta$. Suppose also that the MLE is such that the Cramer regularity conditions are fulfilled, and $\hat{\theta}_n$ is asymptotically normal with mean $\theta$ and variance equal to inverse of the Fisher information matrix. This convergence to normality is convergence in distribution, which does not imply convergence of moments. And yet this MLE is considered asymptotically unbiased (which presumably means that $E[\hat{\theta}_n]\to\theta$ as $n\to\infty$). How does one make this transition from asymptotic normality (with mean equal to $\theta$) to asymptotic unbiasedness (in the sense that $E[\hat{\theta}_n]\to\theta$ as $n\to\infty$)?

Thanks.

$\endgroup$

1 Answer 1

1
$\begingroup$

Adding a moment condition such as $E((\hat\theta_n)^2)\leqslant C$, indeed $E(\hat\theta_n)\to\theta$ follows.

$\endgroup$
2
  • $\begingroup$ I see, thanks. Has anyone verified this condition for the MLEs of the two-parameter Weibull distribution? There are papers dealing with the biases of these MLEs, but how do they know that the expectations of thees MLEs actually converge to the respective parameters in the first place? $\endgroup$
    – Jason
    Oct 21, 2014 at 13:17
  • $\begingroup$ New problem? Then new question? $\endgroup$
    – Did
    Oct 21, 2014 at 13:58

You must log in to answer this question.

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