Need help with finding unbiased estimator

162 Views Asked by At

Suppose that $T$, the time to failure of a component is exponentially distributed. That is, the pdf for $T$ is $${f(x; θ) = \frac{1} {θ} e ^{\textbf{-}x/θ} , 0 ≤ x < ∞}$$. Suppose we test n components and record the failure times $T_1, . . . , T_n$.

(a) Show that $\hatθ = \overline{T} = n^{-1}\sum_{i=1}^{\infty} T_i$ is an unbiased estimator of $θ$.

(b) What is the variance of $\hatθ$?

(c) It can be shown that $Z = \min(T_1, . . . , T_n)$ also has an exponential distribution with parameter $θ/n$. That is $$g(z; θ) = \frac{1} {θ/n}e^ {−z/(θ/n)}$$ (Do not show this). Use this to find another unbiased estimator $\tilde{θ}$ of $θ$ and determine its variance.

For part (a): I showed $E (\hatθ) = E(\theta)$ through rules regarding addition of expected values.

For part (b): Through independence, I showed that $var(\hatθ) = \frac{θ^2}{n}$ (Here I assumed independence which I hope is correct. I'm also not entirely sure that I've done this question correctly).

But I'm having trouble with part (c). I know how to show something is an unbiased estimator but I'm unsure how to find an unbiased estimator. Would appreciate any help.

1

There are 1 best solutions below

1
On BEST ANSWER

For part (c), you can apply Maximum Likelihood Method to the only data , namely $z=\min(t_1,\cdots,t_n)$. Then you will get $\hat{\theta}=nZ$, which is unbiased.