What's the difference between a one-tailed and a two-tailed test? When would you use each?
StatisticsLevel: IntermediateStatistics and probability
What's the difference between a one-tailed and a two-tailed test? When would you use each?
Hints
- Identify the random variable, sample and estimated parameter.
- State the assumptions; distinguish an estimate from its uncertainty.
Answer
A two-sided test rejects for sufficiently large deviations in either direction, for example H₁: μ ≠ μ₀. A pre-specified upper one-sided test rejects only for sufficiently large positive deviations and targets H₁: μ > μ₀. With the same significance level, concentrating the rejection region in one tail can increase power for effects in that chosen direction.
It does not provide a corresponding negative-direction rejection rule. That is different from saying its rejection probability is literally zero for all negative effects: sampling noise can still land in the upper tail. Choose direction before seeing results, based on the decision being made. Do not switch tails after inspecting the estimate; monitor potential harm separately when using a directional success test.
Source and licence: Ishan Dutta — Awesome Data Science Interview Q&A · MIT · answer: Quantoros editorial