Does studying more actually raise your grade? Does eating ice cream really make people drown? Before you compute a single statistic, there is one honest thing to do first: plot the data and look. A scatter plot is the tool for that. Each observation becomes a single dot, placed by its two measurements, and the whole cloud of dots shows — at a glance — whether the two things move together.
It is the most truthful picture in statistics because nothing is hidden. An average throws away every individual and leaves one number; a scatter plot keeps everybody on the page. You see the trend and the exceptions and the weird lone point all at once, before any formula gets a chance to smooth them over.
A scatter plot shows how two variables move together. Each observation contributes one
pair
One dot per observation — so a class of
Three questions read straight off the picture — always ask them in this order.
Next you will put a single number on strength and direction — the
Flip the control to compare a positive cloud (rising to the right), a negative cloud (falling to the right), and a cloud with no association (a shapeless scatter). The same axes hold all three, so only the tilt of the dots changes.
Describing a scatter is a skill you can practise on words alone. Work these through before you look at the answers.
1. A clean trend. Ten towns are plotted with hours of sunshine across and ice-cream sales up. The dots rise steadily from bottom-left to top-right and sit close to a straight line. Read it off: the direction is positive, the form is linear, and because the cloud is tight the relationship is strong. More sun, more ice cream — a strong, positive, linear association.
2. A pattern a straight line would miss. A gardener plots amount of fertiliser across and crop yield up. Yield climbs as fertiliser increases, peaks in the middle, then falls when there is too much. The dots trace an upside-down U. A single straight-line summary through this would look nearly flat and report "no relationship" — completely missing a strong, orderly curved pattern. The lesson: form comes before any line-fitting, because the wrong summary can hide a real effect.
3. One point that shouts. Nine houses show a tidy positive scatter of floor area versus price. A tenth dot sits far off on its own — a tiny cottage selling for a fortune (it came with a beach). This lone outlier, sitting far from the trend, can drag any fitted line toward itself and distort every summary number. Spotting it on the scatter is exactly why you plot first: you decide, with eyes open, whether it belongs.
A scatter can look like a strong pattern that isn't really there — or hide one that is. Two classic traps:
In 1973 the statistician Francis Anscombe built four little datasets — now famous as Anscombe's quartet — that share almost identical means, variances, correlations, and even the same fitted straight line. By the numbers they are twins. But plot them and they are nothing alike: one is a clean straight line, one is a smooth curve, one is a perfect line ruined by a single runaway outlier, and one is a vertical stack of points with one lone dot setting the whole slope. Same summary numbers, four completely different stories.
Anscombe's point, made unforgettable: you must plot your data — never trust summary numbers alone. The modern encore is the "Datasaurus dozen": a dozen datasets, one of them shaped like an actual dinosaur, that all share the same means, standard deviations, and correlation as a plain boring blob. The stats agree; the pictures could not be more different. Plot first, compute second.