SEb in AP Statistics

SEb is the standard error of the slope, an estimate of the standard deviation of the sampling distribution of the slope b. In AP Stats the relevant current topics, it sets the margin of error for a slope confidence interval, b ± t*(SEb), and is calculated as SE = s/(sx√(n-1)).

What is SEb?

SEb answers one question. If you took a new random sample and refit the regression line, how much would the slope b typically change? The slope you calculate is just an estimate of the true population slope β, and SEb measures the typical distance between b and β across all possible samples.

The formula is SE = s/(sx√(n-1)), where s estimates σ (the standard deviation of the responses around the population regression line) and sx is the standard deviation of the x values. Read the formula like a story. More scatter around the line (bigger s) makes the slope wobblier, so SEb grows. More spread in your x values (bigger sx) and a bigger sample size both pin the line down, so SEb shrinks. In practice you almost never compute SEb by hand on the AP exam. It sits in the computer output, in the "SE Coef" column of the row for your explanatory variable, and your job is to find it and use it in b ± t*(SEb).

Why SEb matters in AP® Statistics

SEb lives in Unit 5 (Regression Analysis) and is the engine of the relevant current topics. Learning objective AP Stats the relevant learning objective asks you to determine the margin of error for a slope, which is literally t* times SEb. Then AP Stats the relevant learning objective has you build the full confidence interval, b ± t*(SEb), where b is the point estimate and t*(SEb) is the margin of error. Without SEb there is no interval, no margin of error, and no way to say how precisely you've estimated the true slope β. It also carries into the relevant current topics, since the test statistic for a slope hypothesis test divides by SEb. If you can locate SEb in regression output, you've unlocked most of Unit 5.

How SEb connects across the course

Margin of Error (Units 6-9)

SEb is the building block of the slope's margin of error. Margin of error = t* × SEb, the same critical-value-times-standard-error pattern you used for proportions in Unit 3 and means in Unit 4. Unit 5 just swaps in a new standard error.

Confidence Interval (Units 6-9)

Every CI on the AP exam follows estimate ± (critical value)(standard error). For slopes that's b ± t*(SEb). If you understood one-sample t intervals, you already understand this one; only the standard error formula changed.

Degrees of Freedom (Units 7-9)

The t* that multiplies SEb comes from a t distribution with n - 2 degrees of freedom, not n - 1. Regression estimates two parameters (slope and intercept), so you lose two degrees of freedom. Mixing up n - 1 and n - 2 is a classic point-loser.

Residual Analysis (Units 2 and 9)

SEb is only trustworthy if the regression conditions hold, and you check linearity and constant variance with a residual plot, the same tool from Unit 2. The s inside the SEb formula is built from those residuals.

Is SEb on the AP® Statistics exam?

Multiple choice questions hand you b, SEb, and n, then ask you to construct the interval, find the right t* (remember df = n - 2), or work backward. Fiveable practice questions do exactly this, like computing a 95% interval from b = 3.2 and SEb = 0.8 with 25 students, or backing out t* when you're told a 95% interval has width 5.28. Notice that width = 2 × t* × SEb, so half the width is the margin of error. On FRQs, regression inference appears with computer output, as in released exams like 2021 Question 4, and you have to pull SEb from the correct row (the explanatory variable's row, not the constant's row) and use it in the interval or test statistic. You also have to interpret the result. A wide interval means SEb was large and the slope estimate is imprecise.

SEb vs s (standard deviation of the residuals)

Both appear in regression output, but they measure different wobble. s measures how far individual y values scatter around the regression line, the typical prediction error. SEb measures how much the slope itself would vary from sample to sample. They're linked by SE = s/(sx√(n-1)), so s feeds into SEb, but on the exam you must grab SEb (the slope row's SE column), not s, when building b ± t*(SEb).

Key things to remember about SEb

• SEb estimates the standard deviation of the sampling distribution of the slope b, so it tells you how much the slope would typically change from sample to sample.

• The confidence interval for the true slope β is b ± t*(SEb), where t* comes from a t distribution with n - 2 degrees of freedom.

• The margin of error for the slope is t* × SEb, so if you know the width of an interval, half the width divided by SEb gives you t*.

• The formula SE = s/(sx√(n-1)) means SEb shrinks when you have more data, more spread in x, or less scatter around the line.

• On the exam, find SEb in computer output in the SE column of the explanatory variable's row, not the constant (intercept) row.

• SEb is only meaningful if the conditions check out: linearity, constant variance, independence, and approximately normal responses for each x.

Frequently asked questions about SEb

What is SEb in AP Stats?

SEb is the standard error of the slope, an estimate of how much the sample slope b varies from sample to sample around the true slope β. It's the standard error you plug into the slope confidence interval b ± t*(SEb) in the relevant current topics.

Do I have to calculate SEb by hand on the AP exam?

Almost never. The formula SE = s/(sx√(n-1)) is on the formula sheet, but exam questions nearly always give you SEb in computer output or in the question stem. Your job is to find it and use it correctly, like a question giving b = 3.2 and SEb = 0.8 and asking for the 95% interval.

How is SEb different from s in regression output?

s is the standard deviation of the residuals, measuring how far individual points scatter around the line. SEb measures uncertainty in the slope itself. SEb is built from s through SE = s/(sx√(n-1)), but only SEb goes into b ± t*(SEb).

Is the degrees of freedom for SEb n - 1 or n - 2?

n - 2. Regression estimates two parameters, the slope and the intercept, so the t* multiplying SEb uses n - 2 degrees of freedom. With n = 18 data points, you'd look up t* with 16 degrees of freedom.

Does a small SEb mean the slope is statistically significant?

Not by itself. A small SEb means a precise slope estimate, but significance depends on the ratio b/SEb (the t statistic) and whether the resulting confidence interval contains 0. A small SEb with a slope near 0 can still be non-significant.