ABSTRACT
In Chapter 2, we looked at the theory of building a taxonomy, how one might be created, how to assess whether a taxonomy was reliable, and why reliability mattered. In Chapter 3, we looked a bit more at the subject of reliability and discussed two views of probability: frequentist and Bayesian. Bayesian (or subjective) probability allows people to have prior ideas about what will happen (i.e., a subjective opinion on the probability of any given event; the opinion is then tested against later evidence). We argued that common reliability coefficients, based on frequentist ideas, are problematic, and that reporting simple conditional probabilities (sensitivities and specificities) is less contentious and describes better what we want to know. (We also think formal Bayesian statistics should be considered but recognize that learning how to use these is more time consuming.) In the simplest of terms, we think probability is a profoundly human concept used by and for human beings for pragmatic reasons.
