base rate

Terms from Statistics for HCI: Making Sense of Quantitative Data

The base rate is the underlying level of prevalence of some attribute of a population. For example, the base rate of a disease is the proportion of people who normally have the illness. Where such a base rate is known it can be used as prior probability in Bayesian reasoning. The base rate is important in assessing how likely the underlying attribute is for a particular individual. If you ignore the base rate you may make a poor assessment of the situation, sometimes called the base rate fallacy or base rate neglect. For example, imagine a test for a rare form of cancer that is 100% accurate in detecting if you the cancer and 99% accurate when you don't. You find you have tested positive. At first this seems very worrying; however, if the base rate is that only 1 in a million people have the cancer but the other 999,999 don't, then for every person who has the cancer and tests positive there are 1% of 999,999, that is around 10,000, who test positive and are in the clear. It may be worth having further tests, but you are no more likely to have the disease than to be killed on the road that year.