When a beverage manufacturer checks whether containers meet a 12-ounce target, a sample may differ from the target simply through sampling variation. A hypothesis test frames the decision using two competing statements: the null hypothesis, H₀, and the alternative hypothesis, H₁. The alternative specifies the effect, difference, or relationship that would challenge the null. section 4.1.1 section 4.2.1
The hypotheses are chosen before examining the sample. Their form determines the direction of the test and constrains the conclusions it can support. A composite alternative covers more than one possible parameter value; it may include departures on either side of a target or only departures in one direction. section 4.1.1 section 4.2.1 section 4.2.3
Fill volumes: departures in either direction
For a target mean of 12 ounces, set H₀: μ = 12 and H₁: μ ≠ 12, where μ is the population mean fill volume. The null specifies one value; the alternative includes all means above or below it. Because both underfilling and overfilling matter, the test is two-tailed. section 4.2.1
Recovery rates: improvement in one direction
For a drug evaluated against a historical recovery rate of 50%, set H₀: p = 0.50 and H₁: p > 0.50, where p is the recovery probability. The alternative includes all probabilities above 0.50. Since the question concerns improvement, the test is one-tailed. section 4.2.3
The null is the reference point for evaluating evidence. A p-value measures how unusual the observed result, or a more extreme one, would be if the null were true; it is compared with a predetermined significance level α. Rejecting H₀ supports the alternative under the test procedure, but the decision is made with controlled risk, not certainty. Failing to reject H₀ does not prove it true. section 4.1.2 section 10.2.3 section 4.2.3
The choice of α reflects the consequences of error. A Type I error rejects a true null; a Type II error fails to reject a false null. A smaller α reduces the risk of a false positive but generally increases the risk of a false negative, so the threshold should be chosen deliberately for the problem. section 4.1.2



