Interpret the correlation coefficient, and separate correlation from causation, including lurking variables.
STAT.CORRELATION“I can interpret a correlation coefficient and explain why it is not causation.”
Given r = 0.87, interpret it, then give a plausible lurking-variable explanation for a real correlation.
| Course | Unit | Depth | Role | Checkpoint |
|---|---|---|---|---|
| Algebra II (Grade 10-11) | Probability and statistical inference | intro | introduce | ✓ |
Nothing in the atlas depends on this yet.
A cloud of points with a line the learner drags, showing residuals and total error live.
Least squares becomes an optimisation the learner feels.
engine: plot scene + draggable line + residual segments
Scene archetype: scatter_fit
STAT.CORR.CAUSEConcludes causation from a strong correlation.
Repair: Generate a third-variable explanation for every claim.
One row per instructional system. Rows marked cluster carry authored treatment for this family of topics; rows marked template are derived from the system’s general pattern and are not topic-specific research. Confidence is recorded on every row.
Common Core places correlation and its limits at a specific grade and expects it to be built on a named prerequisite from the year before rather than re-taught from scratch. The standard is usually phrased as understanding plus application, so the expectation is both a correct procedure and an explanation of why it works, developed through a documented progression of representations.
Singapore introduces correlation and its limits concretely with objects the learner can move, then moves to a drawn model — usually a bar or a number bond — that keeps the structure visible, and only then to symbols. The drawn model is retained as the tool for word problems rather than discarded once the algorithm appears, which is why the same bar picture reappears years later for ratio and algebra.
The Russian tradition treats correlation and its limits as structure to be analysed rather than a procedure to be executed. A typical lesson opens with a problem whose condition is schematised — a labelled segment drawing or a quantity table — so the relationships are visible before arithmetic begins, and closes with variations that break any pattern-matching the learner may have adopted. Non-routine and multi-step versions appear early rather than as extension.
Kumon reaches correlation and its limits as a numbered worksheet level rather than a grade, and teaches it by graded repetition: the first pages of the level are barely harder than the last pages of the level before, so the method is inferred rather than explained. There is no manipulative and usually no context — the learner meets the bare computation, repeated until it is both accurate and fast, and repeats the level if the standard completion time is not met.
A Japanese lesson on correlation and its limits usually opens with one problem the class has not been shown how to do, worked independently for a stretch, after which several student approaches are put on the board side by side and ordered from the concrete to the general. The teaching happens in that comparison rather than before it, and the board keeps the whole argument visible so the class can see why the efficient method is efficient.
The classical American approach introduces correlation and its limits as a small increment inside a lesson that also reviews a dozen earlier ideas: the teacher shows a worked example, the learner imitates it, and the topic then reappears in mixed practice for months afterwards. Retention comes from distributed review rather than from a single deep unit, and fact fluency is drilled to automaticity on a timer before the topic is used elsewhere.
Montessori presents correlation and its limits first as a material the child manipulates, chosen so that the structure of the mathematics is physically present in the object rather than explained about it. The child works with that material until the answer becomes predictable, then moves to a deliberately more abstract material representing the same idea, and only reaches written notation when the material has become redundant.
AoPS approaches correlation and its limits by handing the learner a problem that the standard method would solve easily but that they have not yet been given the standard method for, and letting the method be reconstructed from the attempt. The treatment goes materially deeper than a grade-level curriculum, favours a structural argument over a computation, and follows up with non-routine variations chosen to defeat pattern-matching.
The English mastery approach teaches correlation and its limits to the whole class in small steps, using a carefully varied sequence of examples in which one feature changes at a time so the learner can see what the idea does and does not depend on. Concrete and pictorial representations — usually a bar model or a part-whole diagram — sit alongside the symbols rather than before them, and pupils who finish early get deeper problems on the same topic rather than the next one.
IM builds correlation and its limits out of a task students can start with what they already have, then names the mathematics in a whole-class synthesis once several approaches are on the table. Representations are introduced in a planned order within the unit — commonly a diagram, then a table, then the symbolic form — and a short cool-down at the end of the lesson checks whether the idea landed before the sequence moves on.