Let $X_1$ and $X_2$ be independent, normal distributed random variables with equal mean $\mu$ but non-equal standard deviations $\sigma_1$ and $\sigma_2$.

Suppose I know $\sigma_1$ and $\sigma_2$ and I have $n$ samples $x_{11},\ldots,x_{1n}$ from $X_1$ and m samples $x_{21}, \ldots, x_{2m}$ from $X_2$, what is the best estimator for $\mu$? What is its distribution?

(edit: I'm mainly interested in the n=1, m=1 case)

  • 1
    $\begingroup$ Is this homework? $\endgroup$
    – John
    Sep 6, 2011 at 21:12
  • 1
    $\begingroup$ This question for the case $n=1$, $m=1$ is answered at stats.stackexchange.com/questions/12251/… . The reply shows how to obtain the answer in general. $\endgroup$
    – whuber
    Sep 6, 2011 at 21:40
  • $\begingroup$ @John: I need it at my job - why? $\endgroup$ Sep 7, 2011 at 5:12

1 Answer 1


This is analogous to fixed-effect meta-analysis. The best (minimum-variance unbiased) estimator would be the inverse-variance weighted mean $$\hat{\mu} = \frac{ \frac{1}{\sigma_1^2} \sum x_{1i} + \frac{1}{\sigma_2^2} \sum x_{2i} }{n/\sigma_1^2 + m/\sigma_2^2}$$

When $n=m=1$ that reduces to $$\hat{\mu} = \frac{ x_{11} / \sigma_1^2 + x_{21} / \sigma_2^2 }{1/\sigma_1^2 + 1/\sigma_2^2}$$ The distribution of $\hat\mu$ is normal with mean $\mu$ and variance $\frac{ 1 }{n/\sigma_1^2 + m/\sigma_2^2}$.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.