Bonferroni correction

Terms from Statistics for HCI: Making Sense of Quantitative Data

A method of dividing the target p-level by the number of tests, used when there have been multiple tests to reduce the chance of thinking that random effects are real. For example, if you run 10 tests there is an approximately 50:50 chance of seeing at least one come out at 5% significance level by chance. In order to have a real 5% level overall, you therefore test each one at 0.5% (5%/10). It is a form of multiplicity adjustment.

Defined on page 80

Used on pages 80, 92