The population slope (β) is the parameter in the population regression line μy = α + βx that gives the true average change in the response variable y for each one-unit increase in the explanatory variable x. Since you almost never see the whole population, you estimate β using the sample slope b.
The population slope, written as β (beta), is the slope of the population regression line μy = α + βx. It tells you the true average change in the response variable y for every one-unit increase in the explanatory variable x, across the entire population. It's a parameter, a fixed number that exists out there in the world, even though you'll basically never get to calculate it directly.
That's where inference comes in. When you run a regression on sample data, the line you get, ŷ = a + bx, is just an estimate of the population line. The sample slope b is your point estimate for β, and because b changes from sample to sample, you build a confidence interval, b ± t*(SEb), to capture the plausible values of β. Think of it like the relationship between x̄ and μ from Unit 7, except now the unknown parameter is a slope instead of a mean.
The population slope is the whole reason Unit 9 (Inference for Quantitative Data: Slopes) exists. Topic 9.2 is built around estimating β. Learning objective AP Stats 9.2.A asks you to recognize that ŷ = a + bx estimates μy = α + βx, AP Stats 9.2.B has you verify the conditions (linearity, equal standard deviation of y across x, independence, approximately normal responses), AP Stats 9.2.C covers the margin of error t*(SEb), and AP Stats 9.2.D puts it together into the full interval b ± t*(SEb). Every interpretation sentence you write in this unit, like "we are 95% confident that the interval captures the true slope," is a sentence about β. If you confuse the parameter β with the statistic b, your interpretations fall apart, and interpretation is exactly what the exam grades.
Regression Coefficient / Sample Slope b (Units 2 & 9)
The sample slope b from Unit 2's least-squares regression is the point estimate of β. Unit 2 teaches you to compute and interpret b; Unit 9 teaches you to admit that b has sampling variability and to wrap a confidence interval around it.
Confidence Interval (Units 6-9)
The interval for β follows the exact same recipe you learned for proportions and means, point estimate ± (critical value)(standard error). Same logic, new parameter. If you can interpret an interval for μ, you can interpret one for β by swapping in 'true slope.'
Margin of Error (Units 6-9)
For a slope, the margin of error is t*(SEb), where SEb = s/(sx√(n-1)). A practice question with b = 0.35, SE = 0.12, and n = 25 wants exactly this: t* from n - 2 = 23 degrees of freedom times 0.12.
Constant Variance & Conditions for Inference (Unit 9)
You can only build a trustworthy interval for β if the conditions hold, and residual plots are your checking tool. A residual plot with no pattern supports linearity, and roughly even vertical spread supports the equal-standard-deviation condition.
Multiple-choice questions test β in three main ways. First, interpretation: given an interval like (0.35, 0.85), pick the answer that says we're 90% confident the interval captures the true (population) slope, not a statement about b or about 90% of data points. Second, computation: given b, SEb, and n, find the margin of error or the full interval, remembering that degrees of freedom = n - 2. Third, conclusions: an interval like (1.2, 2.8) that excludes 0 gives convincing evidence of a linear relationship between x and y. On the FRQ side, slope inference shows up in Question 5 or 6 territory, where you'd need to name the procedure (t-interval for the slope of a regression line), check all four conditions, compute the interval, and interpret it in context. The grader is looking for the words "true slope" or "population slope," and context for both variables.
β is the parameter, the true slope of the population regression line μy = α + βx, a fixed but unknown number. b is the statistic, the slope of the least-squares line ŷ = a + bx computed from your sample, and it varies from sample to sample. You know b exactly; you estimate β. Confidence intervals are statements about β, never about b. Saying "we're 95% confident the interval captures b" is wrong because you already know b, it's the center of the interval.
The population slope β is the true average change in y for each one-unit increase in x across the whole population, and it's a fixed parameter you estimate rather than calculate.
The sample slope b from the least-squares line ŷ = a + bx is the point estimate of β, and the population line is written μy = α + βx.
The confidence interval for β is b ± t*(SEb), with t* based on n - 2 degrees of freedom.
Before building the interval, check the conditions: the relationship is linear, the standard deviation of y is the same for all x, observations are independent (random sample, 10% condition), and responses are approximately normal for each x.
Interpret the interval as confidence that it captures the true population slope, in context, and never as a statement about the sample slope b.
If the confidence interval for β does not contain 0, you have convincing evidence of a linear relationship between x and y.
It's β, the slope of the population regression line μy = α + βx. It represents the true average change in the response variable y for each one-unit increase in the explanatory variable x across the entire population.
b is the sample slope, a statistic calculated from your data using least squares, while β is the population slope, the unknown parameter you're trying to estimate. b is the point estimate of β, and the interval b ± t*(SEb) gives the plausible values of β.
Almost never. β would require data on the entire population, which is why Unit 9 exists. You estimate β with a confidence interval or test a claim about it with a t-test for slope.
n - 2, not n - 1. A regression line estimates two quantities (slope and intercept), so you lose two degrees of freedom. With n = 25 observations, you'd find t* using df = 23.
Not exactly. It means 0 is a plausible value for β, so you lack convincing evidence of a linear relationship between x and y. If the interval excludes 0, like (1.2, 2.8), that's evidence the true slope isn't zero and a linear relationship likely exists.