01 / 21

Do Those Points
Align?

Explore scatterplots, correlation, and linear regression to discover how data points reveal hidden relationships.

AP Statistics

What Is a Scatterplot?

A scatterplot displays the relationship between two quantitative variables. Each point represents one observation: the horizontal axis shows the explanatory variable \( x \), and the vertical axis shows the response variable \( y \).

Hours Studied Exam Score

Describing a Relationship

Direction
Positive: as \( x \) increases, \( y \) increases. Negative: as \( x \) increases, \( y \) decreases.
Form
Linear: points cluster around a line. Nonlinear: points follow a curve or other shape.
Strength
Strong: points lie close to a clear pattern. Weak: points are widely scattered.

Why Does Alignment Matter?

When data points align closely along a line, we can model the relationship with a linear equation and predict values we haven't observed. The tighter the alignment, the more reliable our predictions.

Strong Weak

The Correlation Coefficient \( r \)

The correlation coefficient \( r \) measures the strength and direction of the linear relationship between two quantitative variables.

\( r \) near +1
Strong positive linear association
\( r \) near 0
No linear association (may still be nonlinear)
\( r \) near −1
Strong negative linear association

Key Properties of \( r \)

1
Bounded: \( -1 \leq r \leq 1 \). Values outside this range are impossible.
2
Unit-free: \( r \) has no units: it stays the same whether you measure in cm or inches.
3
Symmetric: The correlation of \( x \) with \( y \) equals the correlation of \( y \) with \( x \).
4
Sensitive to outliers: A single extreme point can dramatically change \( r \).
5
Linear only: \( r \) measures only straight-line relationships, not curves.

Correlation vs. Causation

A strong correlation does not prove that one variable causes changes in the other. Hidden lurking variables or coincidence may explain the association.

Correlation
Two variables move together. We observe an association, but we do not know why.
Causation
Changing \( x \) directly produces a change in \( y \). Requires a controlled experiment to establish.

The Least-Squares Regression Line

The least-squares regression line (LSRL) is the unique line that minimizes the sum of the squared vertical distances (residuals) from each data point to the line.

\( b \) : Slope
For each 1-unit increase in \( x \), the predicted \( \hat{y} \) changes by \( b \).
\( a \) : Intercept
The predicted \( \hat{y} \) when \( x = 0 \). May not always have practical meaning.

Understanding Residuals

A residual is the difference between an observed value and its predicted value from the regression line:

\( x \)
\( y \)
residual residual
Above line (\( y > \hat{y} \))
Below line (\( y < \hat{y} \))

The Coefficient of Determination \( r^2 \)

\( r^2 \) tells us the proportion of variation in \( y \) that is explained by the linear relationship with \( x \).

\( r^2 = 0.85 \)
85% of the variation in \( y \) is explained by the regression on \( x \).
\( r^2 = 0.25 \)
Only 25% explained: a weak model with lots of unexplained variation.

How to Analyze a Relationship

1
Plot the data: Make a scatterplot with the explanatory variable on the \( x \)-axis.
2
Describe the pattern: State the direction, form, and strength, and note outliers.
3
Compute \( r \): Quantify the linear strength (use technology or the formula).
4
Fit the LSRL: Calculate \( \hat{y} = a + bx \) and interpret slope and intercept in context.
5
Check residuals: Plot residuals vs. \( x \). A random scatter confirms the linear model is appropriate.
6
Report \( r^2 \): Interpret the proportion of variation explained.

Example 1: Study Hours vs. Score

A teacher records study hours (\( x \)) and test scores (\( y \)) for 8 students. The data give \( \bar{x} = 5 \), \( s_x = 2 \), \( \bar{y} = 78 \), \( s_y = 10 \), and \( r = 0.92 \). Find the LSRL.

1
Slope: \( b = r \cdot \dfrac{s_y}{s_x} = 0.92 \cdot \dfrac{10}{2} = 4.6 \)
2
Intercept: \( a = \bar{y} - b\bar{x} = 78 - 4.6(5) = 55 \)
3
Equation: \( \hat{y} = 55 + 4.6x \)

Example 2: Prediction and Residual

Using the LSRL \( \hat{y} = 55 + 4.6x \), predict the score for a student who studied 7 hours. The student actually scored 90. Find the residual.

1
Prediction: \( \hat{y} = 55 + 4.6(7) = 55 + 32.2 = 87.2 \)
2
Residual: \( y - \hat{y} = 90 - 87.2 = 2.8 \)

Example 3: Interpreting \( r^2 \)

Since \( r = 0.92 \), we compute \( r^2 = (0.92)^2 = 0.8464 \).

84.64% Explained 15.36% Total Variation in Test Scores

The remaining 15.36% is due to other factors not captured by study hours alone.

Common Mistakes

✕ Confusing \( r \) and \( r^2 \)
\( r = 0.7 \) does not mean 70% explained. You must square it: \( r^2 = 0.49 \), so only 49% is explained.
✕ Claiming Causation
A high correlation doesn't prove cause. Always consider lurking variables and say "association" not "causes."
✕ Extrapolating
Using the LSRL to predict far beyond the observed data range is unreliable. The pattern may not continue.

Summary Table

Concept Symbol What It Tells You
Correlation \( r \) Strength and direction of linear association
Coefficient of Determination \( r^2 \) % of variation in \( y \) explained by \( x \)
Slope \( b \) Predicted change in \( y \) per 1-unit increase in \( x \)
Intercept \( a \) Predicted \( y \) when \( x = 0 \)
Residual \( y - \hat{y} \) How far an observed value is from the prediction
LSRL \( \hat{y} = a + bx \) Best-fit line minimizing squared residuals

Question 1 of 5

A dataset has \( r = -0.85 \). Which of the following best describes the scatterplot?

Question 2 of 5

If the LSRL is \( \hat{y} = 12 + 3.5x \), what is the predicted value when \( x = 4 \)?

Question 3 of 5

A student's actual score is 88 and the regression line predicts 92. What is the residual?

Question 4 of 5

If \( r = 0.60 \), what percentage of the variation in the response variable is explained by the linear model?

Question 5 of 5

Which statement is always true about the least-squares regression line?