Scatter Diagrams: Reading Correlation from Plots Guide

A scatter diagram (scatter plot) is simply a graph where each observation becomes one dot with coordinates (x, y). By plotting pairs of values — study hours against marks, advertising spend against sales, temperature against ice-cream sales — you can literally see the relationship between two variables before computing a single number.

This visual step matters because numbers alone can mislead. Two datasets can share the same correlation coefficient yet look completely different — one a tidy line, the other a curve, the third a cloud with one wild outlier. The scatter plot shows you which story you are dealing with.

Key Concepts

• Positive correlation: dots drift upward left to right — as x grows, y tends to grow.
• Negative correlation: dots drift downward — as x grows, y tends to shrink.
• No correlation: a shapeless cloud — x tells you nothing about y.
• Strength: a tight, narrow band means strong correlation; a wide cloud means weak correlation.
• Form: the pattern may be linear (straight band) or curved (for example, diminishing returns).
• Outliers: lone dots far from the pack that can distort every statistic you compute afterwards.
• Correlation ≠ causation: ice-cream sales and drownings correlate strongly — both are caused by hot weather, not by each other.

How It Works

1. Choose axes sensibly: put the explanatory variable (the one you think drives things) on the x-axis and the response on the y-axis.
2. Read the direction: upward drift = positive association, downward = negative, none = no association.
3. Judge the strength: imagine an ellipse around the dots — thinner ellipse, stronger correlation.
4. Spot outliers and clusters: mark any dot far from the rest, and check for subgroups.
5. Only then compute: calculate the correlation coefficient r (−1 to +1) knowing what pattern it is summarising.

Worked Example

Five students report (study hours, marks): (2, 45), (4, 55), (5, 62), (7, 74), (8, 80).

1. Plot the dots: they march steadily upward from bottom-left to top-right.
2. Direction and form: clearly positive and roughly linear — more hours, more marks, in a straight-ish band.
3. Strength: the dots hug an imaginary line closely, suggesting a strong correlation (r would come out near 0.99).
4. Rough prediction: a student studying 6 hours should score around 68 — read straight up from x = 6 to the band, then across. Interpolation between (5, 62) and (7, 74) gives (62 + 74) / 2 = 68.
5. Caution: predicting for 20 hours would be extrapolation beyond the data — the pattern may not continue.

Common Misconceptions

• Correlation does not prove causation. A strong r only says two variables move together; a hidden third variable may drive both.
• r = 0 does not mean "no relationship". It means no linear relationship — a perfect U-shaped curve also gives r ≈ 0. The scatter plot reveals what r hides.
• Outliers can fake or hide correlation. One extreme dot can create a strong r out of nothing, or destroy a real one — always look before you compute.

Real-World Applications

• Economics: plotting income against spending to see consumption patterns across households.
• Medicine: dose-response plots showing how a drug's effect changes with dosage.
• Sports: training load versus performance to find the sweet spot before overtraining.
• Business: advertising spend versus revenue to check whether extra ads still pay off.

A scatter diagram is the honest first look at any two-variable relationship: direction, form, strength, and outliers, all visible at a glance. Plot before you calculate, and your statistics will describe reality instead of hiding it.