One-tailed binomial test by direct evaluation
The makers of a puzzle app claim that 30% of players complete the bonus level. A reviewer suspects that the true percentage is lower. She chooses a random sample of 20 players and finds that 2 of them complete the bonus level.
(a) State an assumption that is needed for a binomial distribution to be a suitable model for the number of players in the sample who complete the bonus level.
(b) Test the reviewer's suspicion at the 5% significance level.
- 1
(a) The assumption must be about the players in context, not a restatement of . Either independence or a constant probability earns the mark.
- 2
(b) Let be the population proportion of players who complete the bonus level. The suspicion is that it is lower, so the test is one-tailed (lower tail).
- 3
Under , . For a lower-tail test you need the tail probability , not . Write out the full binomial expression — the method mark depends on seeing it.
- 4
Evaluate each term and add.
- 5
Compare the tail probability with the significance level.
- 6
Conclude in context, without sounding certain: the data give evidence, they do not prove anything.
Answer
(a) Players complete the level independently (or with a constant probability). (b) , reject : there is sufficient evidence to suggest that the percentage of players completing the bonus level is less than 30%.
Examiner tip
The March 2025 examiner report stressed: "It was important that the full Binomial expression was shown here." A bare calculator value of earns only a special-case mark, and alone scores nothing because a valid comparison must use a tail probability. "The makers are wrong" is too definite for the final mark.