How can I use the delta method to show that

$\sqrt n$ (1/$\bar Y_n$ - 1/$\mu)$ $\rightarrow$ N (0, $\sigma^2$/$\mu^4$) ?

We know that:

$\bar Y_n$ = $1/n \sum_{n=1}^n Y_i$

$\ Y_i$ is i.i.d; $E(\ Y_i )= \mu\ne 0$ and $V(\ Y_i)= \sigma^2 \gt$ 0


Set $f(X) = \frac{1}{X}$ (for $f\colon W\longrightarrow \mathbb R$). Then $Df(X) = -\frac{1}{X^2}$ ($Df$ denotes the derivative). The delta-method states that $$\sqrt n(f(X) - f(\mu)) \rightarrow N(0,\sigma^2Df(\mu)^2)$$ provided that $\sqrt n(X-\mu)\rightarrow N(0,\sigma^2)$. Now your job is to plug in the right value for $X$ and you are done ;-)

  • $\begingroup$ Is that sufficient as a proof? $\endgroup$ – Bonsaibubble Jan 9 '17 at 22:34
  • $\begingroup$ I would say yes. $\endgroup$ – Syd Amerikaner Jan 11 '17 at 11:21
  • $\begingroup$ Ok, I was just wondering why you did not apply Slutsky as one of the steps. $\endgroup$ – Bonsaibubble Jan 11 '17 at 14:06
  • $\begingroup$ It's not necessary because we are allowed to use the delta method (however, delta method requires that $\sqrt n(X-\mu)$ converges in distribution. If this is not given, you have to proof it first of course, but that's a new question) $\endgroup$ – Syd Amerikaner Jan 11 '17 at 14:13
  • $\begingroup$ Ah, ok! I thought I have to show that as well $\endgroup$ – Bonsaibubble Jan 11 '17 at 14:38

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.