Skip to content

MODULE 4 · LESSON 22 · Data visualization

Scatterplots

Inspect paired observations, clusters and unusual points before summarizing a numerical relationship.

By Laurence Svekis · Self-paced · All worked data is synthetic unless you supply your own source.

Try the practice ↓

PredictInspectCalculateCheckExplainTransfer

Your starting point

Prerequisites: Use the ideas and vocabulary from lesson 21; revisit its worked examples if needed.

Definition: A scatterplot places each paired observation at an x and y position. Its pattern can show direction, form, clusters and unusual points. Each point should retain the pairing of measurements from the same unit or occasion.

Read the structure before the coefficient

Look for an upward or downward pattern, a curve, separate groups and points far from the rest. A single correlation coefficient cannot summarize every shape. A U-shaped pattern can have little linear correlation while still containing a strong non-linear relationship. Count the valid pairs and remember that overlapping points may hide repeated observations.

Check scales and pairing

Axis units influence interpretation. Hours versus minutes changes the numerical slope, though the same positive linear scaling does not change Pearson correlation. Sorting x and y separately destroys real pairings and can manufacture an orderly trend. If one member of a pair is missing, omit that pair from the plotted relationship and disclose the number excluded.

Interpret groups and influential cases

Different buildings, channels or learner backgrounds may form clusters. A combined pattern can differ from within-group patterns because group composition matters. Inspect an unusual point against the source before removal. A scatterplot suggests questions to investigate; it does not establish the direction of causation, the absence of confounding or the reliability of an extrapolation beyond the observed range.

Two worked examples

Pairing must stay intact

Three fictional learners have pairs (1 hour, 70), (2 hours, 50) and (3 hours, 80). Sorting scores independently into 50,70,80 and reattaching them to hours changes the observations. The resulting upward pattern is partly an artifact of the transformation, not evidence about the learners.

A curved relationship

Points (−2,4), (−1,1), (0,0), (1,1), (2,4) form a U. The positive and negative sides balance in the linear correlation, but the plot clearly shows a systematic curve. Reporting “no relationship” from a near-zero linear coefficient would discard the visible nonlinear structure.

Where this becomes useful

An energy analyst compares outdoor temperature and electricity use. Office and library records may have different baselines because of opening hours or building size. Inspect the paired plot and the groups before proposing an efficiency intervention.

Make an attempt

Use Correlation Investigation to enter the U-shaped pairs from the example. Compare the coefficient with the plotted shape. Then use the Energy sample in the workbench to draw temperature against kwh and explain a plausible grouping effect.

Open the accompanying interactive lab →

Misconception check

Claim: A scatterplot trend tells you which variable causes the other.

Inspect the correction

Direction on the page is not causal direction. Consider common causes, reverse causation and the study design.

Independent application

Construct a paired dataset with two groups or a curve. Plot the original pairs, calculate linear correlation and explain what the coefficient misses. Document the units, missing-pair policy and one causal claim the plot cannot support.

Completion records practice and self-review. It is not certification of statistical competence.