Répartition aléatoire dans les groupes → les conclusions de cause à effet sont permises.
Sélection aléatoire de l'échantillon → les résultats se généralisent à la population.
Étude observationnelle → association seulement.
Méthode pas à pas
Cherche deux choses : y a-t-il eu répartition aléatoire ? y a-t-il eu sélection aléatoire ?
Pièges fréquents
Généraliser les résultats de volontaires à « toutes les personnes ».
Exemple corrigé
Expérience avec répartition aléatoire de volontaires : “the app likely improves sleep for people similar to those studied” (pas “for everyone”).
Exemples de questions
Essayez de répondre avant d'ouvrir le corrigé. Les questions sont dans la langue de l'examen.
Question 1
Difficulté: facile≈ 1 min
A newspaper wants to know what fraction of a city’s residents support building a new stadium. Which of the following methods would most likely produce a biased sample?
AA. Surveying residents chosen at random from a list of all households
BB. Surveying residents selected at random from the city’s tax records
CC. Surveying every 100th name in the city’s telephone directory
DD. Surveying fans leaving a baseball game at the current stadium
Voir la réponse et l'explication
Bonne réponse: D. Surveying fans leaving a baseball game at the current stadium
Explication
Fans at a game are far more likely than average residents to support a stadium, so that sample is biased.
Random selection from household lists or tax records, and systematic selection from a directory, are much more representative.
Question 2
Difficulté: moyenne≈ 1 min
A teacher wants to estimate the mean time her 150 students spend on homework each night. She asks the 25 students in her first-period class. Which population can her result most appropriately be generalized to?
AA. All students in the school
BB. All 150 of her students
CC. Students in her first-period class only
DD. All high school students in the district
Voir la réponse et l'explication
Bonne réponse: C. Students in her first-period class only
Explication
The 25 students were not randomly selected from the 150; they are one intact class, which may differ from the others (time of day, course level). Without random sampling, the result describes only the students surveyed.
Generalizing to the school or district is even less justified.
Question 3
Difficulté: difficile≈ 1 min
A school district compares test scores at two schools. School A’s mean score is 78 and School B’s is 84. A board member concludes that School B’s teachers are more effective. Which of the following, if true, most weakens this conclusion?
AA. School B has newer computers and larger classrooms than School A, which makes teaching there easier.
BB. School B admits students by an entrance exam, while School A admits all students in its neighborhood.
CC. Both schools gave the test on the same day in the same week of the school year, under identical conditions.
DD. School A has considerably more students than School B, so its teachers have less time for each one.
Voir la réponse et l'explication
Bonne réponse: B. School B admits students by an entrance exam, while School A admits all students in its neighborhood.
Explication
If School B selects students by exam, its higher scores may reflect the students it admits rather than its teaching. Student selection is a confounding variable.
Newer computers might even support the teaching explanation; the test date and enrollment size do not explain the gap.