Suppose $X_1, \ldots X_n \sim N(\mu, \sigma^2)$ and we only observe $(X_1 - \bar{X}, \ldots, X_n - \bar{X})$, such that each observed point is now mean centered. Can we still learn about $\mu$ or $\sigma^2$?


Since the variables are de-meaned, I don't think there is a way for you to get back the mean. However, you can make inferences about the variance since (by the way I'm assuming your samples are uncorrelated):

\begin{align} \operatorname{var}(X_i-\bar{X})&=\operatorname{var}\left(\frac{n-1}{n}X_i-\frac{1}{n}\sum_{i\neq j} X_j\right) \\&=\frac{(n-1)^2}{n^2} \operatorname{var}(X_i) + \frac{1}{n^2} \sum_{i\neq j} \operatorname{var}(X_j) \\&=\sigma^2\frac{(n-1)^2}{n^2}+\sigma^2\frac{(n-1)}{n^2}\\ &=\frac{n-1}{n}\sigma^2 \end{align}

  • 6
    $\begingroup$ +1. One rigorous way to support your statement about the mean being unidentifiable would be to observe that $\mu$ doesn't enter into the likelihood at all. $\endgroup$ – whuber Apr 4 at 13:56

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.