Fishing for significance

Also written: fishing, p-hacking

Trying many predictors, interactions or outlier rules until something gets p < 0.05. Even pure random columns will look significant by chance, so decide what to test beforehand and report everything.