Reading a Scatterplot Without Predicting

A table of paired values just lists numbers. A scatterplot turns each pair into a dot, and once you have a whole cloud of dots, your eye catches patterns a list of numbers buries: does y rise or fall as x grows, do the points hug a path or scatter wide, and is that path straight or curved. That's the whole reading, in three quick words: direction, strength, form. Direction is positive, negative, or none. Strength is how tightly the points cling to a trend, from strong to weak. Form is the shape the trend makes, usually linear here, meaning it traces a straight path rather than curving.

These three readings are independent of each other, which is the trap. A cloud can be tight and still show no trend at all, if it's just a round blob sitting in the middle of the grid with no slant. Tightness alone doesn't mean strong linear association; strength only counts the pull toward a trend, and a blob has no trend to pull toward. So a packed cluster with no direction is a weak association, full stop, even though the points sit close together.

This reading stops short of using the line of best fit to forecast a y-value for a new x, that's a separate skill. Here you're just describing the cloud itself, including simple counts: every point sits above the line or below it, never both, so those two counts always add up to the total number of points plotted.