ABSTRACT
All sciences have to deal with uncertainty in hypotheses and theories. If you subscribe to the Bayesian or subjectivist theory of probability, you can talk of these in terms of probabilities. In a statistical science like psychology, you have the additional problem of sampling error. Everything we measure is noisy and laden with error. If I want to measure your simple reaction time or your IQ or how long you take to solve an eight letter anagram, I will likely get a different number every time I try. I might therefore average a number of observations to estimate the ‘true score’ that underlies all the observed scores you are capable of generating. Of course, true scores vary between individuals as well. If I want to know the mean IQ of psychology students, for example, I have to worry about whether the sample of people I choose to test is large enough, or representative enough to give me an accurate answer. This is why psychologists have developed their toolbox of inferential statistical techniques.
