Given $X_1,...,X_n$ ~ i.i.d. $N(\mu, \sigma^2)$ where the mean is unknown, let the estimator for $\sigma^2$ be $\hat{e} = p\sum_{i=1}^n(X_i-\overline{X})^2$ How do I choose $p$ so that this estimator has the minimal mean squared error (MSE)?

I've found that $E(\hat{e}) = p\sigma^2(n-1)$

$Var(\hat{e}) = 2p^2\sigma^4(n-1)$

But I can't find the minimal square error, because everytime I try to solve for the min of $MSE = Bias^2 + Var$, it doesn't work out.


Assuming your calculations of the expectation and variance of the estimator $$\hat e = p \sum_{i=1}^n (X_i - \bar X)^2$$ are correct, then you have $$MSE = \operatorname{E}[\hat e - \sigma^2]^2 + \operatorname{Var}[\hat e] = (p(n-1) - 1)^2 (\sigma^2)^2 + 2p^2 \sigma^4 (n-1),$$ or $$MSE = \sigma^4\left((n^2-1)p^2 - 2(n-1)p + 1\right).$$ Then take the derivative with respect to $p$, set it to zero, and solve for the critical point: this gives $$p = (n+1)^{-1}.$$


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.