# Minimal sufficient statistic whose dimension is less than dimension of parameter

Consider following example:

Suppose $X\sim N(0, \sigma^2)$, consider a random sample of size one from this population. Clearly $X$ is sufficient statistic but $|X|$ is minimal sufficient statistic, which is also complete.

This example shows that if a sufficient statistic has same dimension as complete sufficient statistic then sufficient statistic may not be complete.

I wonder if there exist an example where minimal sufficient statistic has dimension is less than dimension of parameter.

## 1 Answer

An example where the minimal sufficient statistic has dimension less than the dimension of parameter: a single observation from $\rm{Beta}(\alpha,\beta)$-distribution.

When the sample size is only 1 and the parameter dimensionality is large, what can you do? You put your head down and wait for more data to come.

• Is the minimal sufficient statistic x itself? – Tan Nov 7 '20 at 3:15
• @Tan: Yes, it is – kjetil b halvorsen Jan 29 at 16:40