Unbiased estimator of mean of exponential distribution

1.4k Views Asked by At

$X_1,X_2, .. ,X_n$ is a random sample of an exponential distribution with mean $\theta$. Show $nX_{(1)}$ is an unbiased estimator of $\theta$

When I calculated $X_{(1)}$, there's a $n$ in the numerator. There's no way $nX_{(1)}$ is unbiased since there's a $n^2$ in the numerator..

Can someone check if the problem is correct and give me links for the relevant formulas (if the problem is correct)?

1

There are 1 best solutions below

0
On BEST ANSWER

The distribution of the first (minimum) order statistic $X_{(1)} = \min(X_1, \ldots, X_n)$ of an iid sample $(X_1, \ldots X_n)$ drawn from an exponential distribution with mean $\theta$ may be computed as follows. Consider the survival function of $X_{(1)}$, namely $$\begin{align*} S_{X_{(1)}}(x) &= \Pr[X_{(1)} > x] \\ &= \Pr\left[ \bigcap_{i=1}^n (X_i > x) \right] \\ &\overset{\text{ind}}{=} \prod_{i=1}^n \Pr[X_i > x] \\ &\overset{\text{id}}{=} \Pr[X > x]^n \\ &= (e^{-x/\theta})^n \\ &= e^{-nx/\theta}. \end{align*}$$ The equality with 'ind' on top holds because the $X_i$ are independent; the equality with 'id' on top holds because the $X_i$ are identically distributed. Consequently $X_{(1)}$ has probability density $$f_{X_{(1)}}(x) = -S'_{X_{(1)}}(x) = \frac{n}{\theta} e^{-nx/\theta}$$ over the same support as $X$. It is immediately obvious that $X_{(1)}$ is itself exponentially distributed with mean $\theta/n$; therefore, the expectation $$\operatorname{E}[nX_{(1)}] = \theta$$ as claimed.