You take the Unit 5 test in class today. This page is everything you can do before it — what’s on it, the traps, and practice that isn’t on the real test. There are no test questions here.
No new reading. Everything on the test comes from 5.1–5.13. The fastest revision is the 5.13 claim relay, not re-reading. (Want extra practice? Runestone’s 17.9 Exam Questions drill the list-and-string code shapes.)
A checkpoint on the Big Data region. Nothing here is new. After the test: the flex day, then Unit 6.
Where the multiple-choice points are — study to this map:
| What it tests | From | Roughly |
|---|---|---|
| Metadata: what it is, and what it reveals | 5.2 | 2 Q |
| CSV structure: header row, records, fields | 5.1, 5.3 | 1 Q |
| What a program over a dataset computes — count, filter, accumulate | 5.4, 5.7 | 4 Q |
| Cleaning: which values distort a result, and what cleaning changes | 5.6 | 2 Q |
| Chart choice and misleading charts | 5.8 | 3 Q |
| Correlation vs causation; confounding variables | 5.9 | 2 Q |
| Bias, sampling, and missing data | 5.10 | 2 Q |
1. Say the unit in a breath (2 min). Records and fields; clean, filter, group, chart honestly; say only what it supports.
2. Run the six traps (5 min). Below.
3. Redo the claim relay (5 min). 5.13 — until you file every claim right.
4. Take the practice set (8 min). Section 4 — fresh questions, none on the real test.
| Don’t confuse | The one-line difference |
|---|---|
| counting vs summing | count ← count + 1 counts rows; total ← total + x adds values. Read the line inside the IF. |
| exact match vs “looks the same” | To a program, "Brea", "brea", and "Brea " are three different values. |
| placeholder vs real value | 999, -1, or a blank coded as 0 is missing data, not a number. It distorts averages and min/max. |
| accurate vs honest | Correct numbers can still mislead. Check where the axis starts. |
| correlation vs causation | Moving together isn’t proof. Look for the confounder. Pick the most cautious answer that’s still true. |
| big sample vs representative sample | Size doesn’t fix bias. Ask who couldn’t end up in the data. |
Six mixed questions in AP format. None is on the real test — they’re for calibration.
Prompt. A city compares its 40 neighborhoods. Neighborhoods with more miles of bike lanes had fewer car crashes last year. A council member says: “Bike lanes prevent crashes. We should add bike lanes everywhere.”
(a) What does the data support? (b) Name one confounding variable and explain how it could produce this pattern. (c) Name one piece of missing information you would want before deciding. (d) How could acting on this claim go wrong?
□ (a) States exactly what’s supported. More bike-lane miles and fewer crashes went together across these neighborhoods last year. That’s a correlation — not proof that lanes caused it.
□ (b) Names a confounder and links it to BOTH variables. e.g., traffic volume: quieter residential neighborhoods are easier places to add bike lanes and have fewer cars to crash. Naming it without explaining both links earns half.
□ (c) Names specific missing data. Crash counts from before the lanes were built; traffic volume per neighborhood; speed limits; one year vs several.
□ (d) Names a concrete consequence. Lanes added on busy arterial roads might not reduce crashes there — and the city could skip a fix that would have worked, like lower speed limits.
□ Stays cautious without being vague. “More research is needed” alone earns nothing. Every sentence names something.