STAT 401 Exam 3
Wednesday 9:05 to 9:55 · covers Homework 8 and 9 · no calculator, no tables, pencil, one sheet of notes
What the exam will look like
- Source of questions: the class examples and the homework problems, with new numbers and a twist that asks you to explain something in words.
- No tables and no calculator. The numbers will be simple: variances that add to a perfect square, z-scores that land on 1, 2 or 3.
- Confidence intervals are handed to you. You will not compute one from scratch. You interpret it, check the assumptions, and say what changes if the level or the sample size changes.
- Explanations earn the points. Most parts say "explain", "justify" or "interpret". A number alone is not a full answer.
What was said in the review class
Interval given, you interpret. Take the olive oil example. You are told the 95% interval is (472.34, 473.66). Write the interpretation, say whether the assumptions for the method were met, then answer a follow-up.
"My boss wants it more precise, so let's use 99%." That does not make it more precise. A 99% interval from the same data is always wider than the 95% one. You do not recalculate. You say it is wider and explain why.
Interval versus level. The interval is the pair of numbers, from A to B. The level (90, 95, 98, 99) is how confident you are in the method that produced it.
Sample size for a proportion. With no earlier estimate use p = 0.5, because 0.5 × 0.5 is the largest value p(1 − p) can take. If earlier surveys keep giving 0.4, use 0.4 × 0.6. Write the 1068 calculation on your sheet.
Response rate. First find n and round up. Then divide by the response rate and round up again. 1068 at a 50% response rate means you contact 2136.
Sample size for a mean. It is (z × s / margin)², squared, using the standard deviation you are told to use. The margin of error is half the width because the interval is plus and minus.
Multipliers. Know the ones for 90, 95, 98 and 99. You will not be asked for 93 because that needs a table. You may be asked how you would find one.
Empirical rule. 68% within 1 standard deviation, 95% within 2, 99.7% within 3. It needs a bell shape. It works for Z and also for T, because T looks like Z once n is about 30.
Normal approximation to the binomial. You need at least 5 expected successes and at least 5 expected failures. Then move the count by a half: more than 20 is above 20.5, at least 20 is above 19.5, less than 20 is below 19.5, at most 20 is below 20.5.
"Things I say that are simple are not on your sheets." Write the simple things down: the multipliers, 68/95/99.7, the rounding rule, the sentence for an interpretation.
Study order
- 1Read a confidence intervalmost likely
- 2Sample size and response ratemost likely
- 3Multipliers and the empirical rulestressed in review
- 4Binomial to normal with the half-unit correctionstressed in review
- 5Sums, differences, averages and the CLTall of Homework 8
- 6Carry-over from earlier examsExam 2 did this
- 7Question bank: every class example and homework problem
- 8Mock exam, 50 minutes
- 9Exam 2 review: where the questions came from
What Exam 2 taught us
- Every question was a class example or homework problem with a twist. The crack-length question was the dog-lifetime class example with new numbers. The binomial question was the Homework 5 format. The table question was the textbook's own example of two variables that have zero covariance and are still dependent.
- The twist was always a concept. "What single change makes this negative binomial?" "Independent or dependent, and why?" "Which is more likely?" None of these needed a long calculation.
- Old material came back. The five-number summary was from the first unit and it appeared inside a continuous-distribution question.
- So this time: practice the real class examples and homework, write the explanations out in sentences, and keep a carry-over block for earlier units.