I am trying to prove that the variance of the sample lag-1 autocorrelation $$\hat{\rho}=\frac{\sum_{t=1}^n(x_t-\bar{x})(x_{t-1}-\bar{x})}{\sum_{t=1}^n(x_{t-1}-\bar{x})^2}$$ for an i.i.d. R.V is asymptotically normal with mean $0$ and variance $\frac{1}{n}$. The first two assertions are easy, but I am having trouble showing that the variance is $\frac{1}{n}$.

I have tried using a Taylor expansion on the above expression about the actual variance and covariance and then taking the variance of that expression. My taylor expantion looks like:


where SS1 is the sample autovariance, SS2 is the sample variance, $$\sigma_1^2$$ is the actual autocovarince, and $$\sigma_2^2$$ is the actual variance. Using the fact that $$\rho=\frac{SS1}{SS2}$$ I was able to get that the variance is


but this is not the same as $\frac{1}{T}$. My professor has told me that there is a way to prove it using the martingale difference clt and the law of large numbers, but I have not been able to see how that approach would work.

Could someone please help? Thanks, Paul

EDIT: I figured out how to do it the other way, using the martingale difference clt. If we split the summation into $$\hat{\rho}=\frac{\sum_{t=1}^n(x_t-\bar{x})(x_{t-1}-\bar{x})}{\sigma^2}\frac{\sigma^2}{{\sum_{t=1}^n(x_{t-1}-\bar{x})^2}}$$ we can see that the right factor goes to 1 as t->inf. Now we can see that the left factor is a martingale and satisfies the criteria for the Maritingale difference CLT (all relatively easy proofs, though it takes a few step to show that the left factor is a martingale, and you need to use the fact that $$\mathbb{E}(x^2)<\inf$$, an assumption which I didn't mention before).

I am still not sure what went wrong with my Taylor series expansion idea...if the result has instead been $$1/(T*\sigma^2)*(\rho^2+\rho+1)$$ setting $$\rho=0$$ (since x is i.i.d.) would yield the desired result. I will give the answer to anyone who can do it this way.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.