A one-sided, or one-tailed, test places its rejection region in one direction of the test-statistic distribution. It fits a question where only an increase or only a decrease would matter; a question allowing meaningful change in either direction calls for a two-sided test. section 4.2.3 section 4.2.1
Choose the direction from the research question, and state the hypotheses before examining sample data. This helps keep the test from being selected to fit the observed result. The significance level sets the threshold for rejecting the null hypothesis and reflects a trade-off between Type I and Type II errors. section 4.2.3 section 4.2.1
Testing whether a drug improves recovery
With a historical recovery rate of 50%, the hypotheses are H₀: p = 0.50 and H₁: p > 0.50. Because only an increase supports the claim, the example uses a one-tailed exact binomial test for binary patient outcomes. Its null model assumes independent outcomes with the same recovery probability. section 4.2.3

