Asymptotic normality

Also known as: CAN, Asymptotically normal, Consistent and asymptotically normal

An estimator $W_n$ is asymptotically normal if $\sqrt{n}(W_n-\theta)/\sigma_\theta\xrightarrow{d}N(0,1)$ for some asymptotic variance $\sigma_\theta^2$. Together with consistency, this is the CAN property.

The CLT gives asymptotic normality of $\bar X$. Regular MLEs are typically CAN, enabling Wald intervals and tests from an estimated $\widehat{\mathrm{s.e.}}$.

← Back to CS Glossary