Why do I use n - 2 degrees of freedom instead of n - 1?

In regression you estimate two parameters from the data — the slope and the intercept — so you lose two degrees of freedom. That leaves df = n - 2. This is different from a one-sample mean interval, which estimates only one parameter and uses df = n - 1.

Where exactly is the standard error of the slope in the output?

Look at the predictor's row (not the Constant row) and read the SE Coef column. That value is SE_b. The Coef column in the same row gives the sample slope b.

What does it mean if 0 is inside my confidence interval for the slope?

If 0 is a plausible value for , you cannot conclude there is a linear relationship between x and y — a slope of 0 means no linear association. If the entire interval is above or below 0, you have convincing evidence of a real (positive or negative) linear relationship.

How do I check the Normal condition without raw data?

On the exam you typically reference a provided residual plot, histogram, or dotplot of residuals and state it shows no strong skew or outliers. If only a stated random sample and clean residual plot are given, cite those. Never claim normality without pointing to evidence.

AP-STATS-9.1-9.3

U9.1 Confidence Interval for Slope

Learn to build and interpret a t-confidence interval for the slope of a population regression line, read standard error from output, and check LINER conditions.

What you'll do in this lesson

A voice-first session with the Crimsora tutor on U9.1 Confidence Interval for Slope, then targeted practice and FRQs — with the tutor adapting to where you get stuck.

Start U9.1 Confidence Interval for Slope free

What this lesson covers

When you fit a least-squares line to sample data, the slope

b

you calculate is just one estimate of the true population slope

\beta

. Collect a different sample and you would get a slightly different slope. Unit 9 is about quantifying that uncertainty. In this lesson you will construct a

t

-based confidence interval for

\beta

, pull the standard error of the slope straight from computer regression output, and verify the LINER conditions that make the whole procedure valid.

Mastering the slope CI now pays off directly in the next lessons, where the same output and conditions power the

t

-test for slope. Get comfortable reading a regression table and you will breeze through both.

The Population Regression Model

AP Statistics distinguishes between the sample line you compute and the true relationship in the population. The population model is

y = \alpha + \beta x + \varepsilon

, where

\alpha

is the true intercept,

\beta

is the true slope, and

\varepsilon

is a random error term with mean 0 and constant standard deviation

\sigma

.

Your data produce the estimated line

\hat{y} = a + bx

, where

b

is the sample slope estimating

\beta

and

a

estimates

\alpha

. Because

b

changes from sample to sample, it has a sampling distribution. When conditions hold, that distribution is centered at

\beta

with a standard deviation estimated by the standard error of the slope,

SE_b

.

The key exam idea: a confidence interval for

\beta

tells you a plausible range for the true rate of change of

y

per unit of

x

, not for any single prediction. Students often confuse the true slope

\beta

(a fixed parameter) with

b

(a statistic). The interval captures

\beta

; the sample gives you

b

. Interpreting the interval as being about

b

is a common scoring error, so always phrase conclusions in terms of the population slope.

Building the Interval

Every confidence interval in AP Stats follows the same skeleton: point estimate plus or minus a margin of error, where the margin is a critical value times a standard error. For the slope:

b \pm t^{*} \cdot SE_b

The point estimate is the sample slope

b

. The critical value

t^{*}

comes from a

t

-distribution with

df = n - 2

(you lose two degrees of freedom estimating both slope and intercept). The standard error

SE_b

measures how much the slope varies from sample to sample.

Piece	Where it comes from
$b$	"Coef" column, predictor row
$SE_b$	"SE Coef" column, predictor row
$df$	$n - 2$
$t^{*}$	$t$ -table at chosen confidence level and $df$

So for 95% confidence with

n = 20

, use

df = 18

and look up

t^{*} \approx 2.101

. Multiply by

SE_b

, then add and subtract from

b

. A frequent mistake is using

z^{*}

instead of

t^{*}

, or using

df = n - 1

. Always subtract 2 for a single-predictor regression.

Reading Regression Output

The AP exam almost always hands you a computer printout rather than raw data. A typical table looks like this:

Predictor	Coef	SE Coef	T	P
Constant	12.40	3.10	4.00	0.001
Hours	2.85	0.62	4.60	0.000

The slope

b = 2.85

is the Coef for the predictor row (here, Hours). The standard error

SE_b = 0.62

is directly below in the SE Coef column. Do not grab the constant row values by mistake.

Below the table you usually see

s

and

R^2

(or R-Sq). Here

s

estimates

\sigma

, the spread of points around the line, and

R^2

is the proportion of variation in

y

explained by

x

. You do not need

s

R^2

to build the slope interval, but knowing what they mean helps you check conditions and interpret context. The sample size

n

is needed for

df

; if it is not printed directly, count the data points or use degrees of freedom given in an ANOVA section (residual

df = n - 2

LINER Conditions and Interpretation

Before trusting the interval, verify the LINER conditions:

Letter	Condition	How to check
L	Linear	Scatterplot shows linear form; residual plot has no curve
I	Independent	Data from random sample or independent observations; $10\%$ rule if sampling without replacement
N	Normal	Residuals roughly normal; histogram/dotplot not strongly skewed
E	Equal variance	Residual plot shows constant scatter (no fan shape)
R	Random	Data come from a random sample or randomized experiment

Once conditions pass, state the interval and interpret it. Correct interpretation template: "We are 95% confident that the interval from ___ to ___ captures the true slope

\beta

, the mean change in

y

for each one-unit increase in

x

."

Also know the confidence-level interpretation: "If we repeated this sampling many times, about 95% of the intervals constructed this way would capture the true slope." A powerful move on the exam: if the entire interval is above 0 (or entirely below 0), you can conclude there is convincing evidence of a real linear relationship, since 0 is not a plausible value for

\beta

Common Exam Traps

Several predictable errors cost points on slope-CI questions. First, using the wrong degrees of freedom: it is

n - 2

, not

n - 1

. Second, interpreting the interval as being about individual predictions or about the sample slope rather than the population slope

\beta

.

Third, forgetting to check conditions or listing them without evidence. Saying "the data are normal" earns nothing; you must reference a graph or a stated random sample. Fourth, misreading output by taking the constant row's SE Coef instead of the predictor row's.

Finally, watch the units and direction. The slope is a rate: "

y

-units per one

x

-unit." When the interval lies entirely on one side of zero, say so and connect it to whether a linear relationship plausibly exists. When zero is inside the interval, you cannot rule out no linear relationship. Being crisp about all five of these points is usually the difference between full and partial credit.

Key terms

Population slope $\beta$ .: The true rate of change of the mean of $y$ per one-unit increase in $x$ in the model $y = \alpha + \beta x + \varepsilon$ ; the parameter the interval estimates.
Sample slope $b$ .: The slope of the least-squares line computed from data; the point estimate of $\beta$ .
Standard error of the slope $SE_b$ .: An estimate of the standard deviation of the sampling distribution of $b$ ; read from the SE Coef column of regression output.
Degrees of freedom.: For single-predictor regression inference, $df = n - 2$ , reflecting estimation of both slope and intercept.
LINER conditions.: Linear, Independent, Normal residuals, Equal variance, Random — the assumptions required for valid slope inference.
$s$ (residual standard deviation).: An estimate of $\sigma$ , the typical size of residuals around the regression line, reported in the output.
$t^{*}$ critical value.: The multiplier from the $t$ -distribution with $n-2$ degrees of freedom corresponding to the chosen confidence level.

Worked example

A researcher records the number of study hours (

x

) and exam scores (

y

) for a random sample of 22 students. Software gives the slope coefficient 2.85 with SE Coef 0.62, and a residual plot with no pattern and constant scatter. Construct and interpret a 95% confidence interval for the slope of the population regression line.

Step 1: Identify the pieces from the output. The sample slope is

b = 2.85

and the standard error is

SE_b = 0.62

.

Step 2: Find degrees of freedom. With

n = 22

df = n - 2 = 20

.

Step 3: Find the critical value. For 95% confidence at

df = 20

t^{*} \approx 2.086

.

Step 4: Compute the margin of error.

ME = t^{*} \cdot SE_b = 2.086 \times 0.62 \approx 1.293

.

Step 5: Build the interval.

b \pm ME = 2.85 \pm 1.293

, giving approximately

(1.557,\ 4.143)

.

Step 6: Check conditions. Random sample is stated (R and I). The residual plot shows no curve (L) and constant scatter (E), and no strong skew is noted (N). Conditions are met.

Step 7: Interpret. We are 95% confident that the interval from about 1.56 to 4.14 captures the true slope

\beta

, the mean increase in exam score for each additional hour studied. Because the entire interval lies above 0, there is convincing evidence of a positive linear relationship between study hours and exam score.

Practice questions

A regression of monthly sales on advertising spending uses

n = 15

observations. Which degrees of freedom and interval structure are correct for a confidence interval for the slope?

$df = 14$ , interval $b \pm z^{*} \cdot SE_b$
$df = 13$ , interval $b \pm t^{*} \cdot SE_b$
$df = 15$ , interval $b \pm t^{*} \cdot SE_b$
$df = 13$ , interval $b \pm z^{*} \cdot SE_b$

Answer: $df = 13$ , interval $b \pm t^{*} \cdot SE_b$

For a single-predictor regression,

df = n - 2 = 15 - 2 = 13

, and slope inference uses the

t

-distribution, not the normal

z

-distribution, because

\sigma

is estimated. The structure is

b \pm t^{*} \cdot SE_b

Regression output for predicting plant height (

y

, cm) from water amount (

x

, mL) based on 12 plants reports a slope of 0.045 with SE Coef 0.018. Conditions are satisfied. Construct a 90% confidence interval for the slope and interpret it in context.

Answer: Approximately $(0.0127,\ 0.0773)$ ; we are 90% confident this interval captures the true mean increase in height per additional mL of water.

With

n = 12

df = 10

and

t^{*} \approx 1.812

for 90% confidence. Margin of error

= 1.812 \times 0.018 \approx 0.0326

. Interval:

0.045 \pm 0.0326 = (0.0124, 0.0776)

(small rounding differences are fine). Interpretation: we are 90% confident the interval captures the true slope

\beta

, the mean change in plant height in cm per additional mL of water. Since the interval is entirely positive, there is evidence of a positive linear relationship.

A student concludes: "We are 95% confident that the sample slope 2.1 falls between 1.4 and 2.8." Explain what is wrong with this interpretation.

Answer: The interval estimates the population slope $\beta$ , not the sample slope, and the sample slope is already known.

The sample slope

b = 2.1

is a computed statistic; it is not uncertain and does not need an interval. A confidence interval estimates the unknown population parameter

\beta

. The correct statement: we are 95% confident the interval from 1.4 to 2.8 captures the true population slope, the mean change in

y

per one-unit change in

x

FAQ

Why do I use $n - 2$ degrees of freedom instead of $n - 1$ ?: In regression you estimate two parameters from the data — the slope and the intercept — so you lose two degrees of freedom. That leaves $df = n - 2$ . This is different from a one-sample mean interval, which estimates only one parameter and uses $df = n - 1$ .
Where exactly is the standard error of the slope in the output?: Look at the predictor's row (not the Constant row) and read the SE Coef column. That value is $SE_b$ . The Coef column in the same row gives the sample slope $b$ .
What does it mean if 0 is inside my confidence interval for the slope?: If 0 is a plausible value for $\beta$ , you cannot conclude there is a linear relationship between $x$ and $y$ — a slope of 0 means no linear association. If the entire interval is above or below 0, you have convincing evidence of a real (positive or negative) linear relationship.
How do I check the Normal condition without raw data?: On the exam you typically reference a provided residual plot, histogram, or dotplot of residuals and state it shows no strong skew or outliers. If only a stated random sample and clean residual plot are given, cite those. Never claim normality without pointing to evidence.

Learn this with a teacher, not a page

The Crimsora tutor teaches U9.1 Confidence Interval for Slope live — explaining on a whiteboard, asking you questions, and adapting to where you get stuck.

Try a live session free