How do I tell the difference between the line of best fit prediction and an actual data point?

Read the question wording. Phrases like "according to the line of best fit" or "predicted by the model" mean you plug into the equation y = mx + b. Phrases like "based on the actual data" or references to a specific plotted point mean you read the dot's coordinates directly from the graph. The two values often differ, and that difference is the residual.

Does a strong association mean one variable causes the other?

No. Association is not causation. Even a tight linear pattern only shows that two variables move together; a third factor could drive both, or the link could be coincidental. The Digital SAT will not ask you to claim causation from a scatterplot alone.

What exactly is a residual and how do I compute it?

A residual is the actual observed y-value minus the value the line predicts at that same x: residual = y_actual - y_predicted. Compute the predicted value by substituting the point's x into the line's equation, then subtract. Positive residuals sit above the line; negative ones sit below.

When is a prediction from the line trustworthy?

Predictions are most reliable when the association is strong and linear and when the x-value falls within the range of the collected data (interpolation). Predicting far outside that range (extrapolation) is risky because the trend may change, sometimes producing impossible values like negative prices or temperatures.

DSAT-3.4

Two-Variable Data: Scatterplots & Lines of Best Fit

Master Digital SAT scatterplots: describe association, read lines of best fit, interpret slope and intercept in context, and judge residuals and extrapolation.

What you'll do in this lesson

A voice-first session with the Crimsora tutor on Two-Variable Data: Scatterplots & Lines of Best Fit, then targeted practice and FRQs — with the tutor adapting to where you get stuck.

Start Two-Variable Data: Scatterplots & Lines of Best Fit free

What this lesson covers

Two-variable data questions show up often on the Digital SAT, usually as a scatterplot with a drawn line of best fit. The good news: these questions reward careful reading more than heavy computation. If you can describe how two variables move together, pull slope and intercept straight from a line's equation, and translate those numbers into real-world meaning, you can lock in points quickly.

This lesson walks through describing association, using the fitted line to make predictions, interpreting slope and intercept in context, and evaluating residuals and the danger of extrapolation. By the end, you'll know exactly what the test is asking when it shows you a cloud of dots and a straight line running through them.

Describing Association in a Scatterplot

A scatterplot displays paired data: each dot has an

x

-value (explanatory variable) and a

y

-value (response variable). Your first job is to describe the association — the pattern in how

y

changes as

x

increases.

Three features matter: direction, form, and strength. Direction is positive (dots rise left to right), negative (dots fall), or none. Form is linear (dots follow a straight-line pattern) or nonlinear (a curve, such as quadratic or exponential). Strength describes how tightly the dots cluster around the pattern — strong means little scatter, weak means lots of scatter.

Feature	What to look for
Direction	Do dots rise (positive) or fall (negative)?
Form	Straight line or a curve?
Strength	Tightly clustered (strong) or spread out (weak)?

A common misconception is confusing strength with slope. A line can be steep but the data weakly associated, or nearly flat but tightly clustered. Slope measures rate of change; strength measures how closely points follow the trend. Also remember: association is not causation. Even a strong linear pattern does not prove that

x

causes

y

. The Digital SAT may ask you to pick the best description of the relationship, so match all three features to the graph.

Reading and Using the Line of Best Fit

The line of best fit (also called the least-squares regression line) is the straight line that comes closest to all the data points, minimizing the total squared vertical distance between points and line. On the Digital SAT it is usually already drawn and its equation is given in the form

y = mx + b

.

To predict a value, substitute a known variable and solve. If the line is

y = 3.2x + 14

and you want the predicted

y

when

x = 10

, compute

y = 3.2(10) + 14 = 46

. To find the

x

that produces a given

y

, set

y

equal and solve for

x

.

Read carefully whether the question wants a value predicted by the line or an actual data point read from the graph. These often differ. A question might say "according to the line of best fit" — use the equation. If it says "based on the data" or points to a specific dot, read the scatterplot directly.

Watch units and scale. Axis labels frequently include units like "thousands of dollars" or "minutes," and gridlines may count by 2, 5, or 25. Misreading the scale is the most common avoidable error. Always check what one gridline represents before you estimate a coordinate.

Interpreting Slope and Intercept in Context

The Digital SAT loves asking what the slope or intercept means in a real situation. For a line

y = mx + b

:

The slope

m

is the predicted change in

y

for each one-unit increase in

x

. Always state it with units: "For each additional hour studied, the predicted score increases by

m

points." A negative slope means

y

decreases as

x

increases.

The

y

-intercept

b

is the predicted value of

y

when

x = 0

. In context this is a starting value or baseline — but only meaningful if

x = 0

falls within a sensible range. Sometimes

x = 0

is impossible (a person of height 0), so the intercept is just a mathematical anchor, not a real quantity.

Symbol	Meaning	Context template
$m$	rate of change	"per one-unit increase in $x$ , $y$ changes by $m$ "
$b$	value at $x=0$	"when $x$ is 0, predicted $y$ is $b$ "

A frequent trap: swapping slope and intercept, or dropping the word "predicted." The line gives estimates, not exact outcomes. Correct answer choices usually say "predicted" or "estimated," and they attach the right units to the right number. Read every answer choice fully — the difference between right and wrong is often a single unit or a reversed variable.

Residuals and the Limits of Prediction

A residual is the difference between an actual observed value and the value predicted by the line:

\text{residual} = y_{\text{actual}} - y_{\text{predicted}}

. A positive residual means the point sits above the line (the line underestimated); a negative residual means the point sits below the line (the line overestimated).

To compute a residual, read the actual

y

from the data point, plug that point's

x

into the line's equation to get the predicted

y

, and subtract. For example, if a point is at

(8, 50)

and the line predicts

y = 3.2(8) + 14 = 39.6

, the residual is

50 - 39.6 = 10.4

— the point is above the line.

Reliability depends on where you predict. Interpolation (predicting within the range of the data) is generally reliable if the association is strong and linear. Extrapolation (predicting far outside the data range) is risky: the linear pattern may not continue, so those predictions can be badly off. If a question asks whether a prediction for

x = 200

is trustworthy when data only span

x = 0

30

, the answer is that it is unreliable because it extrapolates well beyond the observed data. Recognizing this distinction is a favorite Digital SAT concept.

Key terms

Association.: The relationship between two variables in a scatterplot, described by direction (positive/negative), form (linear/nonlinear), and strength (strong/weak).
Line of best fit.: The straight line that minimizes the total squared vertical distance to the data points; used to model the trend and make predictions, written as $y = mx + b$ .
Slope.: The rate of change $m$ in a linear model; the predicted change in $y$ for each one-unit increase in $x$ .
y-intercept.: The value $b$ ; the predicted value of $y$ when $x = 0$ , meaningful only if $x = 0$ is realistic for the data.
Residual.: The observed value minus the predicted value: $y_{\text{actual}} - y_{\text{predicted}}$ ; positive means the point is above the line, negative means below.
Interpolation.: Predicting a value within the range of the observed data, generally reliable for a strong linear pattern.
Extrapolation.: Predicting a value outside the range of the observed data, which is unreliable because the trend may not continue.

Worked example

A scientist plots the temperature of a cooling liquid (in degrees Celsius) against time (in minutes). The line of best fit is

y = -1.8x + 92

, where

x

is time in minutes and

y

is temperature. (a) Interpret the slope and intercept. (b) Predict the temperature at

x = 15

minutes. (c) One measured data point is

(10, 78)

. Find its residual. (d) Is a prediction for

x = 90

reliable if data span 0 to 25 minutes?

Part (a): The slope is

-1.8

, so the liquid's temperature is predicted to decrease by 1.8 degrees Celsius for each additional minute. The intercept is 92, meaning at time

x = 0

(the start), the predicted temperature is 92 degrees Celsius.

Part (b): Substitute

x = 15

into the line:

y = -1.8(15) + 92 = -27 + 92 = 65

. The predicted temperature at 15 minutes is 65 degrees Celsius.

Part (c): First find the predicted value at

x = 10

y = -1.8(10) + 92 = -18 + 92 = 74

. The actual value is 78. The residual is

y_{\text{actual}} - y_{\text{predicted}} = 78 - 74 = 4

. The positive residual means this point lies above the line, so the line underestimated the temperature here.

Part (d): The data only cover 0 to 25 minutes, but

x = 90

is far outside that range. This is extrapolation, so the prediction is unreliable — the cooling likely slows and levels off near room temperature rather than continuing to drop 1.8 degrees per minute, which would give an unrealistic negative temperature.

Practice questions

A scatterplot relating a car's age

x

(in years) to its resale value

y

(in thousands of dollars) has line of best fit

y = -1.4x + 22

. Which statement best interpprets the slope?

For each additional year of age, the predicted resale value decreases by 1.4 thousand dollars.
For each additional year of age, the predicted resale value increases by 1.4 thousand dollars.
When the car is new, its predicted resale value is 1.4 thousand dollars.
For each additional 1.4 years, the resale value decreases by 1 thousand dollars.

Answer: For each additional year of age, the predicted resale value decreases by 1.4 thousand dollars.

The slope

-1.4

is the predicted change in

y

per one-unit increase in

x

. Since

x

is age in years and

y

is value in thousands of dollars, each extra year lowers predicted value by 1.4 thousand dollars. The negative sign signals a decrease. The choice describing the intercept (1.4 when new) confuses slope with intercept, and the reciprocal-rate choice misreads what slope measures.

Using the same line

y = -1.4x + 22

, a 5-year-old car actually sold for 17 thousand dollars. Find the residual and state whether the point lies above or below the line of best fit.

Answer: The residual is $2$ thousand dollars, and the point lies above the line.

Predicted value at

x = 5

y = -1.4(5) + 22 = -7 + 22 = 15

. Residual

= y_{\text{actual}} - y_{\text{predicted}} = 17 - 15 = 2

thousand dollars. A positive residual means the observed value exceeds the prediction, so the point lies above the line. (Always subtract predicted from actual, and keep the units consistent with the

y

-axis.)

The data used to build the line

y = -1.4x + 22

covered cars from 1 to 9 years old. Explain why using the line to predict the resale value of a 30-year-old car may be unreliable.

Answer: Because $x = 30$ is far outside the 1-to-9-year range of the data, the prediction is an extrapolation and the linear trend may not hold that far out.

Predictions are trustworthy mainly within the observed data range (interpolation). At

x = 30

the equation gives

y = -1.4(30) + 22 = -20

, a negative value that is impossible for resale price. This shows the linear model breaks down far beyond the data, illustrating why extrapolated predictions are unreliable.

FAQ

How do I tell the difference between the line of best fit prediction and an actual data point?: Read the question wording. Phrases like "according to the line of best fit" or "predicted by the model" mean you plug into the equation $y = mx + b$ . Phrases like "based on the actual data" or references to a specific plotted point mean you read the dot's coordinates directly from the graph. The two values often differ, and that difference is the residual.
Does a strong association mean one variable causes the other?: No. Association is not causation. Even a tight linear pattern only shows that two variables move together; a third factor could drive both, or the link could be coincidental. The Digital SAT will not ask you to claim causation from a scatterplot alone.
What exactly is a residual and how do I compute it?: A residual is the actual observed $y$ -value minus the value the line predicts at that same $x$ : $\text{residual} = y_{\text{actual}} - y_{\text{predicted}}$ . Compute the predicted value by substituting the point's $x$ into the line's equation, then subtract. Positive residuals sit above the line; negative ones sit below.
When is a prediction from the line trustworthy?: Predictions are most reliable when the association is strong and linear and when the $x$ -value falls within the range of the collected data (interpolation). Predicting far outside that range (extrapolation) is risky because the trend may change, sometimes producing impossible values like negative prices or temperatures.

Learn this with a teacher, not a page

The Crimsora tutor teaches Two-Variable Data: Scatterplots & Lines of Best Fit live — explaining on a whiteboard, asking you questions, and adapting to where you get stuck.

Try a live session free