A small randomized experiment assigns 10 subjects randomly to treatment (n=5) or control (n=5). Treatment scores: 75, 82, 88, 91, 79. Control scores: 70, 72, 68, 75, 71. Mean difference = 10 points. Under randomization inference (Fisher's exact test), the null hypothesis is no causal effect—the observed outcomes are fixed and the randomization is what is random.
Which statements about randomization inference are correct?
A) The p-value equals the fraction of all possible random assignments yielding a difference ≥ 10 under the null
B) Randomization inference requires normally distributed outcomes
C) Randomization inference provides exact inference in finite samples without distributional assumptions
D) Randomization inference can only be applied with equal-sized treatment and control groups
A small randomized experiment assigns 10 subjects randomly to treatment (n=5) or control (n=5). Treatment scores: 75, 82, 88, 91, 79. Control scores: 70, 72, 68, 75, 71. Mean difference = 10 points. Under randomization inference (Fisher's exact test), the null hypothesis is no causal effect—the observed outcomes are fixed and the randomization is what is random.
Which statements about randomization inference are correct?
A) The p-value equals the fraction of all possible random assignments yielding a difference ≥ 10 under the null
B) Randomization inference requires normally distributed outcomes
C) Randomization inference provides exact inference in finite samples without distributional assumptions
D) Randomization inference can only be applied with equal-sized treatment and control groups