Self-Study Sufficient Statistics, Pdf with Indicator Function

292 Views Asked by At

The example is from the Book Hogg Introduction to Mathematical Statistics Page 384, Chapter 7.2 Sufficient Statistics. Please let me know if my argument for the solution is correct, since I used a different one then presented in the book.

Problem:

Let $X_{(1)} < X_{(2)} < \dots < X_{(n)}$ denote the oder statistics of a random sample of sice n from the distribution with pdf:

$$ f(x;\theta) = exp(-(x-\theta)) \mathbb{I}_{[\theta, \infty)} $$

Solution:

By factorization theorem: $L(X;\theta) = g(T(X), \theta)h(x)$:

\begin{align*} L(x;\theta) &= \prod_{i=1}^nexp(-(x_i-\theta)) \mathbb{I}\{\theta < x_i < \infty\} \\ &= exp(-(\sum_{i=1}^nx_i- n\theta)) \cdot \mathbb{I}\{\theta < {x_i}_{(i=1,\dots,n)} < \infty\} \\ &= \mathbb{I}\{max\{X_i\} < \infty\}exp(-\sum_{i=1}^nx_i)\cdot exp(n\theta) \cdot \mathbb{I}\{\theta < min\{X_i\}\} \end{align*}

By factorization theorem:

$$ h(x) = \mathbb{I}\{max\{X_i\} < \infty\}exp(-\sum_{i=1}^nx_i) $$

and: $$ g(T(X), \theta) = exp(n\theta) \cdot \mathbb{I}\{\theta < min\{X_i\}\} \Rightarrow T(X) = min(X) $$

1

There are 1 best solutions below

1
On BEST ANSWER

Yes correct, but in a very simple way

$$f_X(x|\theta)=e^{\theta}e^{-x}\cdot\mathbb{1}_{[\theta;\infty)}(x)$$

Now simple observing that (here $X_1,...,X_n$ are not ordered...)

$\theta \leq X_1<\infty$

$\theta \leq X_2< \infty$

...

$\theta \leq X_n< \infty$

thus it is self evident that

$$\theta\leq X_{(1)}$$

Thus the likelihood is

$$L(\theta)=\underbrace{e^{-\Sigma_iX_i}}_{h(\mathbf{x})}\cdot\underbrace{ e^{n\theta}\cdot\mathbb{1}_{(-\infty;x_{(1)}]}(\theta)}_{g[t(\mathbf{x}),\theta]}$$

thus $T=X_{(1)}$


I edited your question because the order satistic is always written as $X_{(i)}$