A form of cherry picking that tries different kinds of statistical technique till something comes out 'right'. For example, a non-parametric test, followed by a t-test, followed by a Bayesian test, and so on. Sometimes a change is justified, for example if it is clear that the wrong kind of test for the distribution was chosen initially, but the reasons for making such a change must be based on the data, not the results.

Defined on page 81

Used on page 81