Two-Variable Data: Scatterplots and Lines of Best Fit

Two-variable data questions ask how one quantity changes as another one changes — through scatterplots, lines of best fit, and simple linear or exponential models. This guide explains what the SAT actually tests here, works through two full examples using real coordinate data, and covers the vocabulary you need for association and correlation questions.

What Two-Variable Data Means on the SAT

While one-variable data describes a single list of numbers, two-variable data describes the relationship between two related quantities — like hours studied and test score, or advertising spend and monthly sales. The SAT presents this as a scatterplot (a graph of individual data points) and often overlays or asks about a line of best fit, the straight line that most closely models the trend in the data.

Key Vocabulary You Need

• Positive association: as one variable increases, the other tends to increase too (points trend upward left to right).
• Negative association: as one variable increases, the other tends to decrease (points trend downward left to right).
• Line of best fit: a line, usually written as \(y = mx + b\), chosen to minimize overall distance from all the data points.
• Residual: the vertical distance between an actual data point and the value predicted by the line of best fit — a large residual means that point is a poor fit for the model.
• Correlation strength: how tightly the points cluster around the line — tightly clustered points show a strong association; widely scattered points show a weak one.
• Extrapolation: using a line of best fit to predict values well outside the range of the original data — the further you extrapolate, the less reliable the prediction, since the real relationship might not stay linear that far out.

Worked Example: Finding and Using a Line of Best Fit

A researcher records the relationship between weekly study hours and a practice test score for six students:

The line of best fit for this data is approximately \(y = 3x + 62\), where \(x\) is weekly study hours and \(y\) is the predicted score.

Question 1: Using this model, what score does the line predict for a student who studies 7 hours a week?

Step 1 — Substitute \(x = 7\). \(y = 3(7) + 62 = 21 + 62 = 83\).

Question 2: A student who actually studies 8 hours scored 84. What is the residual for this data point?

Step 1 — Find the predicted value at \(x = 8\). \(y = 3(8) + 62 = 86\).

Step 2 — Subtract: actual minus predicted. \(84 - 86 = -2\).

A residual of \(-2\) means the actual score was 2 points below what the line predicted — a small residual, which tells you this point fits the model reasonably well. Watch the order of subtraction: residual is always actual minus predicted, and the sign matters if a question asks whether a point falls above or below the line.

Interpreting Slope and Intercept in Context

SAT questions almost always ask you to explain what the slope and y-intercept mean in real-world terms, not just to identify them. In the equation \(y = 3x + 62\): the slope, 3, means each additional hour of studying is associated with an increase of about 3 points in predicted score. The y-intercept, 62, is the model's predicted score for a student who studies 0 hours per week — useful as a baseline, though it may not be realistic if no student in the data actually studied zero hours.

Worked Example: Negative Association and a Negative Slope

Not every scatterplot trends upward. A negative association is just as common, and the SAT tests slope interpretation the same way regardless of direction. Consider average monthly outdoor temperature and a household's heating bill:

The line of best fit here is \(y = -3x + 280\), where \(x\) is average temperature and \(y\) is predicted heating cost in dollars.

Question 1: What does the slope of \(-3\) mean in context?

Answer: Each 1°F increase in average temperature is associated with a predicted decrease of $3 in monthly heating cost. The negative sign matters — it tells you the association is negative, not that cost itself becomes negative.

Question 2: A sixth household with an average temperature of 90°F actually had a heating cost of $15. Find the residual for this data point.

Step 1 — Find the predicted cost at \(x = 90\). \(y = -3(90) + 280 = -270 + 280 = 10\).

Step 2 — Subtract: actual minus predicted. \(15 - 10 = 5\).

A residual of \(+5\) means the household's actual heating cost was $5 higher than the line predicted — a small, positive residual on an otherwise strongly negative association.

Linear vs. Exponential Models

Occasionally a scatterplot shows a curved pattern that a straight line doesn't fit well — this signals an exponential or quadratic model instead. The tell-tale sign: if the data values are roughly multiplying by a constant factor over equal intervals (rather than adding a constant amount), an exponential model fits better than a line. For instance, if a bacteria population measures 100, 200, 400, 800, and 1,600 at five equally spaced time intervals, each value doubles rather than adding a fixed amount — a clear signal for an exponential model over a linear one, even before you plot a single point. You won't need to derive these models from scratch often, but you do need to recognize when a linear line of best fit is clearly a poor match for curved data.

Frequently asked questions

Do I need to calculate a line of best fit by hand on the SAT?

Rarely from raw data — you're almost always given the equation or the graph and asked to use, interpret, or evaluate it. The embedded Desmos calculator, available throughout the digital SAT Math section, can also plot points if you need to check a trend visually.

Does the SAT test the correlation coefficient r as a number?

No. The digital SAT tests the concept of correlation strength — how tightly points cluster around a line of best fit — described qualitatively as strong, moderate, or weak, and by direction (positive or negative). You won't be asked to calculate a numerical r-value from a formula.

What's the difference between correlation and causation on these questions?

A strong association between two variables doesn't mean one causes the other — this distinction is tested directly in our guide on evaluating statistical claims, which covers observational studies versus experiments.

How is a residual different from an error in measurement?

A residual specifically refers to the gap between an actual data point and the value the line of best fit predicts — it's a property of the model's fit, not a measurement mistake.

Practice interpreting scatterplots and lines of best fit with SAT test series covering this exact subtopic, or continue to inference from sample statistics and margin of error to see how these ideas extend to sampling.