On the Power and Legitimacy of Follow-Up Testing.

Here one should draw a profound distinction between the illegitimate procedure of uncorrected p-hacking on a completed dataset and the valid procedure of post hoc sample incrementing to bolster near-significant results.

also in 2,306 other papers

This analysis shows that it is always advantageous to run follow-up tests for an initial sample that was close to significance, given that the appropriate correction for multiple tests is applied.

also in 2,130 other papers

This prohibition is generally extended to both preliminary data analysis as a study goes along, and post hoc sampling increases when a result tests as marginally significant, by adding a few extra cases to increase the sample size.

also in 26,082 other papers