Cross Validated is a question and answer site for people interested in statistics, machine learning, data analysis, data mining, and data visualization. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

My textbook gives an example, that normal distribution family $\{N(0,\sigma^2):\sigma\in R^+\}$ is not complete, but a complete statistic, $T_n=\sum_{i=1}^n X_i^2$, can still be constructed from samples $(X_1,\cdots,X_n)$.

So under what (sufficient / necessary) conditions can we draw complete statistics from an incomplete model?

share|improve this question
up vote 3 down vote accepted

Necessary conditions might be hard, at least for the question as wide-open as it currently stands. A pretty simple sufficient condition is for $X_{1},X_{2},\ldots,X_{n}$ to be a simple random sample from an exponential family. This covers your example above because the $N(0,\sigma)$ family can be written in the form $$ f(x|\sigma)=\frac{1}{\sqrt{2\pi}}\cdot \frac{1}{\sigma}\cdot \exp\left( \frac{-1}{\sigma^2}\cdot x^2 \right). $$ In one of Lehmann's books (I believe it is Testing Statistical Hypotheses) there is a proof that the canonical statistic $T$ when sampling from an exponential family is complete (plus it's minimal sufficient, ...the list goes on).

Off the top of my head, with necessary conditions you're going on an adventure with image measures, transforms, functional analysis...?

share|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.