3
$\begingroup$

In All of Statistics, chapter 11 (pg. 183), Larry Wasserman states in his description of the Wald Test:

We are testing the null hypothesis $ \hat{\theta} = \theta_0 $ versus the alternative hypothesis $ \theta \neq \theta_0 $.

He then says we should assume that $ \hat{\theta} $ is asymptotically normal, i.e. converges in distribution to a standard normal: $$ \frac{\sqrt{n}(\hat{\theta} - \theta_0)}{\hat{\text{se}}} \rightsquigarrow N(0, 1). $$

(Wasserman uses $ \rightsquigarrow $ to denote converging in distribution.)

My impression is that this is a reasonable assumption because of the Central Limit Theorem.

But then, when proving the following theorem:

Asymptotically the Wald test has size $ \alpha $, that is, $$ \mathbb{P}_{\theta_0}\left(\lvert Z \rvert > z_{\alpha/2}\right) \rightarrow \alpha $$ as $ n \rightarrow \infty $,

He says:

Under $ \theta = \theta_0 $, $ (\hat{\theta} - \theta_0)/\hat{se} \rightsquigarrow N(0, 1). $

I don't see how $ (\hat{\theta} - \theta_0)/\hat{se} \rightsquigarrow N(0, 1) $ follows from the fact that $$ \frac{\sqrt{n}(\hat{\theta} - \theta_0)}{\hat{\text{se}}} \rightsquigarrow N(0, 1). $$

Doesn't this imply that $$ \lim_{n \rightarrow \infty} P\left(\frac{\sqrt{n}(\hat{\theta} - \theta_0)}{\hat{\text{se}}} \leq z\right) = \lim_{n \rightarrow \infty} P\left((\hat{\theta} - \theta_0)/\hat{se} \leq z\right), $$ which is not true?

Is this a mistake or am I missing something?

$\endgroup$
3
  • 1
    $\begingroup$ I cannot find these passages in Chapter 11 of the book. I see the results in Chapter 10, and I don't see the factor of $\sqrt n$ anywhere in the book in defining asymptotic normality (which is good, because if you are normalizing by the standard error then the $\sqrt n$ shouldn't be there). So if you want help you will need to be explicit about the edition of the book you are looking at and where precisely all of these things are defined. $\endgroup$ – guy Apr 3 '20 at 22:45
  • $\begingroup$ So I was using a PDF version that was based on the course notes, but I just downloaded the e-textbook version and noticed that you're right. In section 10.1, the $ \sqrt{n} $ term has been removed and was presumably a mistake. $\endgroup$ – an1lam Apr 3 '20 at 22:55
  • $\begingroup$ @guy: Do you want to leave a short answer just noting that this is wrong because the sample standard error includes the $ \sqrt{n} $ term and I'll mark it as correct so that it's clear this is resolved? $\endgroup$ – an1lam Apr 3 '20 at 22:59
1
$\begingroup$

It’s an error, the sample standard error shouldn’t have square root of $ n $.

$\endgroup$

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.