0
$\begingroup$

Question

Let $X_1,X_2,...,X_n$ be independent continuous random variables with the probability density function, $f(x)= \frac{1}{\mu}e^{-\frac{x}{\mu}}$, if $x>0$ and $0$ otherwise, where $\mu>0$. Find the maximum likelihood estimator for $\mu$ and compute the variance of the maximum likelihood estimator for $\mu$.

Answer

I found the likelihood function to be \begin{equation} l(\mu)=\prod_{i=1}^{n} \frac{1}{\mu}e^{-\frac{X_i}{\mu}}=\frac{1}{\mu^n}e^{-\frac{1}{\mu}\sum_{i=1}^{n}X_i} \end{equation} and thus the log likelihood is \begin{equation} L(\mu)=-n\log_e(\mu)-\frac{1}{\mu}\sum_{i=1}^{n}Xi. \end{equation} So, \begin{equation} \frac{\partial L(\mu)}{\partial \mu}=-\frac{n}{\mu}+\frac{1}{\mu^2}\sum_{i=1}^{n}X_i. \end{equation} Equating this to $0$ gives $\frac{1}{\mu^2}\sum_{i=1}^{n}Xi=\frac{n}{\mu}$ and so $\bar X= \hat \mu$.

However, I am not sure that I wrote out the likelihood function correctly and I don't know how I would find the variance. I know $\text{Var}(X)=E(X^2)-(E(X))^2$ but I don't know how to calculate this using the MLE.

$\endgroup$
1
$\begingroup$

Your work is correct, although personally, I would write the full likelihood as $$\begin{align*} \mathcal L(\mu \mid \boldsymbol x) &= \prod_{i=1}^n \frac{1}{\mu} e^{-x_i/\mu} \mathbb 1 (x_i \ge 0) \\ &= \mu^{-n} \exp\left(-\frac{1}{\mu} \sum_{i=1}^n x_i\right) \mathbb 1 (x_{(1)} \ge 0) \\ &= \mu^{-n} e^{-n \bar x/\mu} \mathbb 1 (x_{(1)} \ge 0), \end{align*}$$ where $\bar x$ is the sample mean; $$\mathbb 1(x_i \ge 0) = \begin{cases} 1, & x_i \ge 0 \\ 0, & \text{otherwise} \end{cases}$$ is an indicator function, and $x_{(1)} = \min_i x_i$ is the first (minimum) order statistic of the sample $\boldsymbol x = (x_1, x_2, \ldots, x_n)$. This however does not affect the validity of your computation of the MLE, which is indeed $\hat \mu = \bar x$, the sample mean. To calculate the variance of this estimator, we write $$\operatorname{Var}[\hat \mu] = \operatorname{Var}[\bar X] = \operatorname{Var}\left[\frac{1}{n} \sum_{i=1}^n X_i\right] \overset{\text{ind}}{=} \frac{1}{n^2} \sum_{i=1}^n \operatorname{Var}[X_i] \overset{\text{id}}{=} \frac{1}{n^2} \cdot n \operatorname{Var}[X] = \frac{\operatorname{Var}[X]}{n},$$ where "ind" over the equality sign indicates that the equality holds because the observations are independent; "id" over the equality sign indicates that equality holds because the observations are identically distributed; and $\operatorname{Var}[X]$ represents the variance of a single observation from this distribution. What is $\operatorname{Var}[X]$ in terms of the parameter $\mu$?

$\endgroup$

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.