check classmates’ panels for correct models, fair comparisons and honest claims, and give specific feedback.
Phase 4: Build
Do Now. Reread one section of the NIST handbook, the part with a warning in its title. Then answer.
According to the NIST handbook, what can a scatter plot never do by itself?
The reading opens over the lesson; close it to come back to exactly where you were.
Inspectors for the truth. Today every team checks other teams’ panels. Check the math (is each model right?) and check the claims (does each sentence say only what the data show?). Use the stems I notice…, I wonder… and Try….
Worked example 1: Rosa compares two models. Team A’s sleep model is the equation y = 6x + 34. Team B shows its model as a table: x = 6, 7, 8, 9 gives y = 71, 76.5, 82, 87.5. Rosa finds Team B’s slope: 76.5 − 71 = 5.5 for each hour. Its intercept: 71 = 5.5(6) + b, so b = 71 − 33 = 38, and Team B’s model is y = 5.5x + 38. Team A’s rate of change is greater (6 vs 5.5 points per hour). At x = 8 both give 82: 6(8) + 34 = 82 and 5.5(8) + 38 = 82. Her note: “I notice your models agree at 8 hours. I wonder which fits the points better at 6 and 9.5 hours. Try counting points above and below each line.”
Worked example 2: Sam checks a claim. A draft panel says: “Phones in bedrooms cause students to lose sleep.” Sam checks the table: 30% vs 60% is a real association in the sample. But the claim says cause. His note: “I notice the percents are correct. I wonder if something else, like late practices, explains both. Try: In our sample, students with phones in the bedroom were less likely to get 8 hours of sleep.”
The checklist. Correct: slopes, intercepts, predictions and percents checked a second way. Clear: title, axes with units, legend, source. Useful: one finding a board member can act on. Honest: sample labels, the model’s range, and no cause claims.
Practice checking before the gallery walk.
Gallery walk. Visit at least two other panels.
For each panel you visit, check one number a second way and one claim for honest wording. Write one I notice, one I wonder and one Try note for each.
Back at your panel, list the feedback you received. Choose the two changes you will make first and explain why they matter most to a school board reader.
This is a sample story, but the pattern is a classic. A team surveys students in grades 1 through 5 at an elementary school. Its scatter plot of shoe size (x) and reading score (y) shows a strong positive linear association. A draft headline says: “Bigger feet make better readers!”
A teammate asks what else changes from grade 1 to grade 5. Age. Older students have bigger feet and have had more years of reading practice. Age is a lurking variable that drives both. When the team makes a separate plot for each grade, NIST’s conditioning-plot idea, the pattern inside each grade almost disappears.
New shoes will not raise anyone’s reading score. The association was real; the cause was not what the headline claimed.
Think about it: What lurking variable might sit behind our Sleep and Scores pattern? How could our panel word its finding honestly?
Our dashboard is only as strong as its most careful sentence.
Exit ticket: rewrite this claim so it is honest: “More sleep makes students get better grades.”
Pacing: warm-up and reading 6 minutes, mini lecture 7, case study 6, guided practice 11, gallery walk 8, closure 2.
Likely misconceptions: treating a strong association as proof; thinking feedback means praise only; comparing models at only one input.
Supports: printed feedback cards with the three stems and the four criteria. Stretch: sketch a conditioning plot for the shoe-size case, one small plot per grade.
Answer guide: warm-up: prove causation; 5.5; both give 82 at 8 hours; not a function; true; “associated”; compare rates. CO2 item: 2125 is x = 110, and 2.6 × 110 + 401 = 286 + 401 = 687, so the arithmetic is right, but 110 years is far outside the 2015-2025 data (extrapolation); try a prediction for a year near the data and add a range warning. Exit ticket: “In our sample, students who slept more tended to have higher quiz scores.” The case study is a sample scenario; the challenge quotation is copied exactly from the NIST page, where the two halves after “even though” are set as indented lines.
National Institute of Standards and Technology, NIST/SEMATECH e-Handbook of Statistical Methods: 1.3.3.26 Scatter Plot · U.S. government work, public domain
The NIST statistics handbook says the heart of science and engineering is discovering cause-and-effect relationships that can be predicted and repeated. A scatter plot is a key tool for that work, because it reveals association: a structured pattern between two variables, such as a straight-line or a curved pattern.
Then the handbook adds a warning, under a heading that says causality is not proved by association. If one thing causes another, the two will be associated. But the reverse is not true: an association does not mean one thing causes the other. The handbook calls association only step one toward cause and effect.
It goes further. No statistical method, the scatter plot included, can prove cause and effect. A plot can strongly suggest a cause. Only a researcher, using knowledge of the science behind the data, can conclude that a cause really exists.
The handbook also describes a conditioning plot, which makes separate scatter plots of Y against X for different values of a third variable. That idea helps when a hidden third variable might be behind a pattern.
“Note, however, that even though causality implies association association does NOT imply causality.”
Source: National Institute of Standards and Technology, NIST/SEMATECH e-Handbook of Statistical Methods: 1.3.3.26 Scatter Plot (itl.nist.gov)