Interpreting your measurements
Much of the early work on experimental design was about creating a methodology that would overcome the personality and prejudice of the person doing the science. The construction of the hypotheses and the clear delineation of the experiment are meant to remove the human aspect – the human error. The classical Latin squares randomisation and blinding were developed to specifically exclude bias in the results, because of the prejudices that might be imported. A paper by the Revd Thomas Bayes was published that argued that the context of an observation was intrinsic to its interpretation. He introduced the idea that prior expectations can be combined with observation, leading to modified or posterior, and expectations. Throughout the twentieth century this Bayesian philosophy ran parallel to the ‘frequentist’ approach of most applied statisticians. The predictive value of the test is P – the probability that the individual is diseased given that they have a positive test.