Which variable goes on the x-axis and which on the y-axis
Every point on a scatter plot is one row of your data. The X column sets how far right it sits and the Y column sets how high. You assign both under Chart columns in the Data tab.
Ask which one you could change on purpose to see what happens to the other. In the preset, a student could study more hours and then see whether the score moves. Hours studied is X and test score is Y. The variable doing the influencing goes on the x-axis, and the one that responds goes on the y-axis. If neither drives the other, like height and arm span, either order is fine.
Swapping the axes does not change the pattern, but it changes the story. Put score on X and hours on Y and the chart quietly suggests that scores cause studying.
Add a line of best fit and R²
A line of best fit is a straight line through the middle of your points that shows the overall direction.
To add one, open the Style tab, expand Scatter options, and turn on Trend line. Chartjuice draws a least-squares line: the one that makes the squared vertical distances from the points as small as possible. Tick Show R² to print the R² value on the chart.
What R² tells you
R² runs from 0 to 1. It is the share of the variation in Y that the line accounts for.
- Near 1, say 0.90: the points hug the line, and X predicts Y well.
- Around 0.5: a clear trend with plenty of scatter around it.
- Near 0: the line explains almost nothing.
R² does not say which way the line slopes. Look at the line for the direction.
Using the line well
One point far from the rest can tilt the line, so check outliers first. Reading a value off the line within your X range is reasonable, but stretching it far past the last point is a guess. If the points bend, R² will be low even though there is a clear pattern, so look at the chart before you trust the number. A line worked out in class or on a graphing calculator uses least squares too, so the two should agree.
Reading a scatter plot: positive, negative, and no correlation
- Positive correlation: points climb from lower left to upper right, like hours studied and test score.
- Negative correlation: points fall from upper left to lower right, like the age of a car and its resale value.
- No correlation: a shapeless blob with no tilt. Knowing X tells you nothing about Y.
A tight band is a strong relationship and a wide, loose band is a weak one.
Far-off points deserve a second look. A typo, a one-off or the most interesting row in your data all look the same here. Two separate clumps often mean two groups mixed together, and one line across both describes neither well.
Correlation isn't causation
Two things moving together does not mean one causes the other. Ice cream sales and sunburns both rise in July, but a cone does not burn anyone. Hot weather drives both. A scatter plot shows that a relationship exists and how strong it is. It cannot show why.
Bubble charts: add a third value with point size
A bubble chart is a scatter plot where each point's size shows a third number. Say each point is a region, X is average income, Y is average rent, and the bubble is the number of survey respondents.
Set the Bubble size column under Chart columns, then turn on Bubbles (size from a column) in the Scatter options. Lower the Opacity so overlapping bubbles stay visible, and try to stay under a few dozen. The bubble chart maker opens with bubbles already turned on.
Color points by group
Put a category in the Group role and each group, such as a class period, gets its own color and a legend entry. Groups take the palette's colors in order; change them in the Colors section of the Style tab.
When a scatter plot is the wrong chart
- Values over time, such as monthly visitors: make a line graph.
- The spread of a single variable, like exam scores: make a histogram.
- Thousands of overlapping points: make a heatmap, which uses color instead of position.
