IS RANDOM FAIR?

The statistician's view of randomness- our useful tool: As statisticians we are well used to explaining how random sampling lies at the heart of the acquisition of new knowledge, and despite its counter-intuitive nature, it is usually a reliable technique. We can add that random sampling has a proven track record, and that there are countless examples [1] of failures where inadequate methods were used. To apply a bit of technical jargon we say that a 'fair ' sample is one which is drawn randomly, that is when each member of the population has an equal (sometimes a measured) chance of being selected. There are mathematical derivations based on the notion of b.l.u.e.- best least-squared unbiased estimators which enable us to work out best estimates of statistics about the population, and also to predict likely errors. Some deviation from this scientific, provable procedure is allowed, for example the use of quota sampling to reduce maximum likely errors at the expense of overall accuracy. What we can say is that random sampling produces a fair result within predictable bounds; that straying from randomness is unreliable and runs the risk of ‘bias’. As statisticians we understand that randomness is benign, a tool for understanding, a device for overcoming our own acknowledged weakness or inability to choose fairly, without bias. Could this idea be used mor