Scores & base rates
Reverse a 1–5 item with 6 − answer. Mean = total ÷ number of scores; median = middle value in order. z = (score − mean) ÷ SD. About 68% of scores are within 1 SD of the mean and 95% within 2.
Square a correlation r for a rough share. With a rare trait, count true flags from the small group and false flags from the big group, then divide true flags by all flags.
One step
Scores, flips, percents and z.Scale score
Six 1–5 items are answered 5, 3, 4, 2, 4 and 3, with none reversed. Find the total and the mean (p.8).
Flip it
On a 1–5 scale, reverse answers of 1, 4 and 5 (p.9).
Retest
In a made-up retest, 18 of 24 people got the same result. What percent is that (p.14)?
z-score
A made-up test has mean 50 and SD 8. Find z for a score of 58.
Two or three steps
Say what each number is compared with.Mixed items
Six 1–5 items are answered 4, 2, 5, 3, 1 and 4. Items 2 and 5 are reversed. Find the total after flipping, and the mean to one decimal place.
Outlier
Seven made-up scores are 11, 14, 15, 15, 17, 18 and 36. Find the mean and median. Then leave out the 36 and find both again. Which changed more?
Bell curve
A made-up scale has mean 100 and SD 15. Between which scores are about 68% of people? About 95%? Of 500 people, about how many are within 1 SD?
Compare
Mean 40, SD 6. Find z for scores of 49 and 34. Which score is about the 16th percentile?
Correlation
One made-up study finds r = 0.4; another finds r = −0.6. Square each for a rough share. Which link is stronger?
Base rate
A made-up test screens 2,000 people; 40 have a rare trait. It catches 75% of them and wrongly flags 5% of the rest. How many are flagged, and what percent of flagged people really have the trait?
Equations
A1–A2 (pp.17–18).Undo the formula
A 10-item scale reports P = (r − 10) ÷ 40 × 100. What raw score r gives 85%?
Two tools
Survey tool A costs 90 + n dollars for n responses; tool B costs 15 + 2.5n. When are they equal, and at what cost?
T-scores
A T-score is T = 50 + 10z. Find z for T = 72. On a made-up test with mean 30 and SD 5, what raw score is that?
Pairs
Every pair in a group compares results once, making n(n − 1) ÷ 2 = 66 pairs. Solve for n.
Your turn to be the test designer
- Pick a number from the book: a scale, a mean, an SD, a retest percent or a base rate.
- Write a question that uses it with a made-up, harmless questionnaire.
- Solve it, and say what the number can’t tell you about a real person.
My problem, and my worked answer:
Try every problem first, then check here. Wrong answers are useful: find the step where your work and the key split apart.
The key never prints with Print my work or Print blank. Open it to get its own Print the answer key button.
Answer key
Worked solutions. Scores, groups, tests and prices are made-up examples unless the book cites them.