I'm wondering if this density has a sufficient statistic to summarize an independent and identically distributed sample of random variables $X_1,X_2,...,X_n$ of this density.

The density is:


After some computation, arrived to:

$f(\textbf{x}|\theta)=e^{n\theta-\sum_{i=1}^{n} x_i}e^{\sum_{i=1}^{n}(-e^{-(x_i-\theta)})}I_{\mathbb{R}^n}(\textbf{x})$

But with this expression, it seems impossible to resume the information in a statistic because of the exponent of the middle factor. There, the $x_i$'s will allways be separated, but together (each one) with the parameter ($\theta$). So it will be impossible to factorize this expression (to use the Halmos Factorization Theorem), meaning that there is no sufficient statistic. Am I concluding correctly, or i'm missing something?


$$ f(\textbf{x}|\theta)=e^{n\theta-\sum_{i=1}^{n} x_i}e^{\sum_{i=1}^{n}(-e^{-(x_i-\theta)})}I_{\mathbb{R}^n}(\textbf{x}) = e^{-\sum_{i=1}^{n} x_i}\cdot e^{n\theta-e^\theta\sum_{i=1}^{n}e^{-x_i}}=h(x)g\left(\theta,\sum_{i=1}^{n}e^{-x_i}\right) $$ By Neyman-Fisher factorization theorem, $T=\sum_{i=1}^{n}e^{-x_i}$ is a sufficient statistic for $\theta$.


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.