Sign up ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

I have a random sample drawn from a $N(\theta,\sigma^2)$ distribution with $\sigma^2$ known. I am trying to estimate $\theta$.

I need to calculate the efficiency of the unbiased estimator, $\bar{X}^2-\frac{\sigma^2}{n}$. To do so, I have been trying to find the pdf of the estimator by using the MGF. However, I have been having some difficulty in doing so.

My thought process thus far:

$M_{\frac{1}{n^2} \sum_{i=1}^{n} X_{i}^{2} -\frac{\sigma^2}{n} (t)} =e^{\frac{-\sigma^2}{n}t}M_{\sum_{i=1}^{n}X_{i}^2}(\frac{1}{n^2}t)$

Can we say that $X_{i}^2=\chi_{1}^2$? If so:

$e^{-\frac{\sigma^2}{n}t}[M_{X_{i}^2}(\frac{1}{n^2}t)]^n =e^{-\frac{\sigma^2}{n}t}[(\frac{1}{1-2(\frac{1}{n^2}t)})^{\frac{1}{2}}]^n =e^{-\frac{\sigma^2}{n}t}[\frac{1}{1-\frac{2}{n^2}}t]^{\frac{n}{2}}$

Do we instead need to take into account that $X\sim N(\theta,\sigma^2)$? Does that mean we need to make $\frac{(X_{i}-\theta)^2}{\sigma^2}$ appear in our equation? If so, how would I do that? Does the following make any sense?

$\frac{1}{n^2}\displaystyle \sum_{i=1}^{n}X_{i}^2-\frac{\sigma^2}{n} =\sigma^2 \frac{1}{\sigma^2}\frac{1}{n^2}\displaystyle \sum_{i=1}^{n}X_{i}^2-\frac{\sigma^2}{n} =\sigma^2\frac{1}{\sigma^2}\frac{1}{n^2}\displaystyle \sum_{i=1}^nX_{i}^2-2\theta X_i+\theta^2+2\theta X_i-\theta^2-\frac{\sigma^2}{n} =\sigma^2\frac{1}{n^2}\displaystyle \sum_{i=1}^{n}\frac{(X_i-\theta)^2}{\sigma^2}+\frac{1}{n^2}\displaystyle \sum_{i=1}^n 2\theta X_i-\frac{1}{n^2}\displaystyle \sum_{i=1}^n \theta^2-\frac{\sigma^2}{n} =\frac{\sigma^2}{n^2}\chi^2_n+\frac{2\theta}{n^2}\displaystyle \sum_{i=1}^{n}X_i-\frac{\theta^2}{n}-\frac{\sigma^2}{n}$

*We'd then take the MGF of the above expression.

Are we now dealing with the sum of a chi-squared and normal distribution? Am I overthinking this?

I'd appreciate any guidance! As you can see, I am quite confused as to what to do. Thank you in advance!

share|cite|improve this question
Oh, I'm sorry! I should have been more clear. The random sample is drawn from a $N(\theta, \sigma^2)$ distribution, where $\sigma^2$ is known. We are estimating $\theta$. –  Jess Feb 5 '13 at 16:30
$\overline{X}$ has perfect efficiency, i.e., it saturates the Cramer-Rao bound. –  zyx Feb 5 '13 at 23:49

1 Answer 1

In order to proof ordinary least squares estimator as BLUE (best linear unbiased estimator) it is sufficient to proof Gauss Markov theorem. I guess you could probably try to go in that direction. A proof if the Gauss Markov theorem can be found under

share|cite|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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