1: \begin{abstract}\noindent%

2: Some practical results are derived for population inference based on a

3: sample, under the two qualitative conditions of `ignorability' and

4: exchangeability.  These are the `Histogram Theorem', for predicting the

5: outcome of a non-sampled member of the population, and its application

6: to inference about the population, both without and with groups.  There

7: are discussions of parametric versus non-parametric models, and different

8: approaches to marginalisation.  An Appendix gives a self-contained proof

9: of the Representation Theorem for finite exchangeable sequences.

10: 

11: \medskip\noindent%

12: \textsc{Keywords}: sampling, surveys, prediction, Finite Representation Theorem

13: \end{abstract}

14: