01 / 22

Confidence Intervals
Slope of Regression Models

Learn to construct and interpret confidence intervals for the true slope \(\beta_1\) of a population regression line using sample data.

AP Statistics

What Is a Regression Slope?

In simple linear regression, we model the relationship between an explanatory variable \(x\) and a response variable \(y\) with the equation:

The slope \(b_1\) tells us the predicted change in \(y\) for each one-unit increase in \(x\). But \(b_1\) is only an estimate of the true population slope \(\beta_1\).

Sample vs. Population Parameters

\(b_1\) : Sample Slope
Calculated from your data. A statistic that varies from sample to sample.
\(\beta_1\) : Population Slope
The true slope of the regression line for the entire population. A fixed but unknown parameter.

Visualizing Sampling Variability

x y True \(\beta_1\) Different samples → different slopes Best fit Sample lines

Each sample produces a slightly different \(b_1\). The dashed lines represent regression lines from different random samples, all estimating the same population slope.

The Confidence Interval Formula

\(b_1\)
Sample slope: the point estimate from your data.
\(t^*\)
Critical value from the \(t\)-distribution with \(df = n - 2\).
\(SE_{b_1}\)
Standard error of the slope, measuring precision of \(b_1\).

Standard Error of \(b_1\)

Where \(s\) is the residual standard deviation (the standard deviation of the residuals), and the denominator captures the spread of the \(x\)-values.

Conditions: L-I-N-E-R

Before constructing a confidence interval for \(\beta_1\), you must verify these conditions:

L
Linear: The scatterplot of \(y\) vs. \(x\) shows a linear pattern (check residual plot for no curve).
I
Independent: Observations are independent (random sample, or \(n < 10\%\) of population).
N
Normal: For any given \(x\), the responses \(y\) are approximately normally distributed (check a histogram or normal probability plot of residuals).
E
Equal variance: The residual plot shows roughly constant spread across all \(x\)-values (no fanning).
R
Random: Data come from a random sample or randomized experiment.

CI vs. Hypothesis Test for \(\beta_1\)

Confidence Interval Hypothesis Test
Goal Estimate a range for \(\beta_1\) Test if \(\beta_1 = 0\)
Formula \(b_1 \pm t^* \cdot SE_{b_1}\) \(t = \frac{b_1 - 0}{SE_{b_1}}\)
Key Output Interval \((L, U)\) \(p\)-value
Connection If 0 is not in the CI → reject \(H_0\!: \beta_1 = 0\)

Reading Computer Output

On the AP exam, you will see a table like this. Here is how to find the values you need:

Predictor Coef SE Coef T P
Constant 12.40 2.15 5.77 0.000
Hours Studied 2.34 0.45 5.20 0.000

How to Construct the Interval

1
Check conditions: Verify L-I-N-E-R using the residual plot, context, and sample information.
2
Identify values: Read \(b_1\) and \(SE_{b_1}\) from the computer output.
3
Find \(t^*\): Use \(df = n - 2\) and the desired confidence level (e.g., 95%) in the \(t\)-table or invT function.
4
Calculate: \(b_1 \pm t^* \cdot SE_{b_1}\). Write both the lower and upper bounds.
5
Interpret in context: "We are C% confident that the true slope of the population regression line relating [y] to [x] is between (L, U)."

Example 1: Study Hours and Exam Scores

A sample of \(n = 20\) students gives \(b_1 = 2.34\) and \(SE_{b_1} = 0.45\). Construct a 95% confidence interval for \(\beta_1\).

1
Degrees of freedom: \(df = 20 - 2 = 18\)
2
Critical value: \(t^* = 2.101\) (from \(t\)-table, \(df = 18\), 95%)
3
Margin of error: \(2.101 \times 0.45 = 0.945\)
4
Interval: \(2.34 \pm 0.945 = (1.395,\ 3.285)\)

Visualizing the Interval

0 1 2 3 4 1.395 3.285 \(b_1 = 2.34\) 0 is outside 95% Confidence Interval for \(\beta_1\)

Example 2: Temperature and Ice Cream Sales

A researcher collects data on daily temperature (\(^{\circ}\text{F}\)) and ice cream sales (\$) for \(n = 30\) days. Computer output shows:

Predictor Coef SE Coef
Constant -15.20 8.73
Temperature 3.85 0.62
1
\(df = 30 - 2 = 28\), and for 99% confidence: \(t^* = 2.763\)
2
ME \(= 2.763 \times 0.62 = 1.713\)
3
99% CI: \(3.85 \pm 1.713 = (2.137,\ 5.563)\)

Example 3: Does 0 Fall in the Interval?

A study with \(n = 15\) participants finds \(b_1 = 0.82\) and \(SE_{b_1} = 0.55\). Build a 95% CI and determine if the linear relationship is significant.

1
\(df = 15 - 2 = 13\), so \(t^* = 2.160\)
2
ME \(= 2.160 \times 0.55 = 1.188\)
3
95% CI: \(0.82 \pm 1.188 = (-0.368,\ 2.008)\)

Interpreting with Precision

A correct AP interpretation must include three elements:

1. Confidence Level
"We are C% confident that..."
2. Parameter
"...the true slope of the population regression line relating \([y]\) to \([x]\)..."
3. Interval
"...is between \(L\) and \(U\)."

Common Mistakes

✕ Wrong df
Using \(df = n - 1\) instead of \(df = n - 2\). Regression uses two estimated parameters (\(b_0\) and \(b_1\)).
✕ Probability Language
Saying "there is a 95% probability \(\beta_1\) is in the interval." Confidence describes the method, not the parameter.
✕ Ignoring Conditions
Not checking linearity, independence, normality, and equal variance before building the CI.
✕ Reading Wrong Row
Using the constant row's SE instead of the predictor row's SE from computer output.

What Affects Interval Width?

Factor Effect on Width Why
Increase \(n\) Narrower Smaller \(SE_{b_1}\) and smaller \(t^*\)
Increase confidence level Wider Larger \(t^*\) critical value
More spread in \(x\) Narrower Larger \(\sum(x_i - \bar{x})^2\) → smaller \(SE_{b_1}\)
More scatter around line Wider Larger \(s\) → larger \(SE_{b_1}\)

Question 1 of 5

A regression analysis with \(n = 25\) observations produces \(b_1 = 4.2\) and \(SE_{b_1} = 1.1\). What are the degrees of freedom for the \(t\)-distribution?

Question 2 of 5

A 95% confidence interval for \(\beta_1\) is \((1.8,\ 5.4)\). Which interpretation is correct?

Question 3 of 5

Computer output for a regression with \(n = 22\) shows \(b_1 = 3.50\) and \(SE_{b_1} = 0.80\). Using \(t^* = 2.086\) for 95% confidence, what is the confidence interval?

Question 4 of 5

A 90% confidence interval for \(\beta_1\) is \((-0.42,\ 1.86)\). What can we conclude about the significance of the linear relationship at \(\alpha = 0.10\)?

Question 5 of 5

Which of the following would decrease the width of a confidence interval for \(\beta_1\)?