# MLE for normal sample with known mean [closed]

What is the Maximum Likelihood Estimator of the variance of a sample $X_1\dots X_n$ of i.i.d variables with normal distribution $X_i\sim \mathcal{N}(\mu_0,\sigma^2)$ when the mean $\mu_0$ is known? Show that this estimator is unbiased.

I've worked out the general case $X_i\sim \mathcal{N}(\mu,\sigma^2)$ with unknown mean $\mu$, using the Likelihood function. In that case, the Maximum Likelihood estimator of the variance turns out to be biased. I'm stuck as in how to use the known value of the mean to convert the estimator into an unbiased one.

## closed as off-topic by C. Falcon, Daniel W. Farlow, Namaste, Leucippus, JunivenMar 15 '17 at 1:00

This question appears to be off-topic. The users who voted to close gave this specific reason:

• "This question is missing context or other details: Please improve the question by providing additional context, which ideally includes your thoughts on the problem and any attempts you have made to solve it. This information helps others identify where you have difficulties and helps them write answers appropriate to your experience level." – C. Falcon, Daniel W. Farlow, Namaste, Leucippus, Juniven
If this question can be reworded to fit the rules in the help center, please edit the question.

• I'm sorry I don't know how to use the math notations here. But I've worked out the general case, using the Likelihoid function. Where the Variance estimator turns out to be biased. I'm stuck as in how to use the known value of the mean to convert the estimator into an unbiased one. – Kriss Mar 14 '17 at 20:15

The log-likelihood of a Gaussian sample $X_1 \dots X_n$ with $X_i\sim \mathcal{N}(\mu_0,\sigma^2)$ is $$L = -\frac{n}{2} \ln(2\pi) - \frac{n}{2} \ln\sigma^2 - \frac{1}{2\sigma^2} \sum_{i=1}^n \left(X_i - \mu_0\right)^2 \, ,$$ where $\sigma^2$ is the parameter to estimate. The derivative of the log-likelihood with respect to $\sigma^2$ vanishes at $\sigma^2 = \hat{\sigma}^2_n$, with $$\hat{\sigma}^2_n = \frac{1}{n} \sum_{i=1}^n \left(X_i - \mu_0\right)^2 \, .$$ One shows that $\sigma^2 = \hat{\sigma}^2_n$ corresponds to a maximum of the log-likelihood. Therefore, $\hat{\sigma}^2_n$ defines the maximum likelihood estimator. Now, we compute the expected value of the estimator $$E(\hat{\sigma}^2_n) = \frac{1}{n} \sum_{i=1}^n E\left(\left(X_i - \mu_0\right)^2\right) .$$ Since $\mu_0 = E(X_i)$, one has $E\left(\left(X_i - \mu_0\right)^2\right) = \sigma^2$ for all $i$. Finally, $E(\hat{\sigma}^2_n) = \sigma^2$. The estimator is unbiased.
For comparison, in the case where the mean is unknown, the log-likelihood $$L = -\frac{n}{2} \ln(2\pi) - \frac{n}{2} \ln\sigma^2 - \frac{1}{2\sigma^2} \sum_{i=1}^n \left(X_i - \mu\right)^2$$ is maximized with respect to the parameter $(\mu,\sigma^2)$. The maximum likelihood estimator is \begin{aligned} &\tilde{\mu}_n = \overline{X}_n = \frac{1}{n} \sum_{i=1}^n X_i\, ,\\ &\tilde{\sigma}^2_n = \frac{1}{n} \sum_{i=1}^n \left(X_i - \overline{X}_n\right)^2\, , \end{aligned} where the variance estimator $\tilde{\sigma}^2_n$ is biased: $E(\tilde{\sigma}^2_n) = \frac{n-1}{n}\sigma^2 \neq \sigma^2$.