AP Statistics Glossary 2027
These are the terms that earn (or lose) FRQ points on the redesigned AP Statistics exam. Each entry pairs a plain definition with the rubric-worthy phrasing graders look for.
Terms are grouped alphabetically. Jump to a term or read straight through. The phrasing under "On FRQs" is the wording most likely to earn (or miss) the rubric point.
Parameter vs. Statistic
A parameter is a number describing a population (μ, p, σ). A statistic is a number describing a sample (x̄, p̂, s). Inference uses statistics to estimate parameters.
On FRQs: Always define your parameter in the actual context of the problem — "μ = mean commute time in minutes for all workers in the city" — not just "μ = the population mean."
Random Sampling vs. Random Assignment
Random sampling is how subjects are selected — it enables generalization to the population. Random assignment is how subjects are placed into treatments — it enables cause-and-effect conclusions.
On FRQs: Scope of inference: sampling → generalize; assignment → causation; both → both; neither → neither. This shows up in almost every experimental-design FRQ subpart.
Confidence Interval
A range of plausible values for a parameter, constructed from sample data at a given confidence level. Wider = less precise; higher confidence = wider.
On FRQs: "95% confident" means the procedure captures the true parameter in 95% of repeated samples — not that there's a 95% chance the parameter falls in this specific interval.
P-value
The probability of observing a test statistic at least as extreme as the one from the sample, assuming H₀ is true.
On FRQs: Never write "the p-value is the probability that H₀ is true." That's the single most-penalized misinterpretation on the exam.
Sampling Distribution
The distribution of a statistic over all possible samples of a given size from a population.
On FRQs: Three distinct distributions: population, sample, sampling. Confusing them costs a point on almost every sampling-distribution FRQ.
Central Limit Theorem
For large samples, the sampling distribution of x̄ is approximately normal, regardless of the population's shape. Rule: n ≥ 30, or the population is normal.
On FRQs: Cite the CLT by name to justify a t-procedure when the population is not stated to be normal.
Type I and Type II Errors
Type I: rejecting a true H₀ (probability α). Type II: failing to reject a false H₀ (probability β). Power = 1 − β.
On FRQs: Describe each error in the context of the problem, not in the abstract. Increasing α raises power but also raises Type I risk.
Inference Conditions
Every AP inference procedure requires three condition checks: Random (sample or assignment), Independence (10% rule when sampling without replacement), and Normal/Large Counts (Large Counts for proportions, or CLT/plot for means).
On FRQs: "Conditions are met" is worth zero points. Check each condition specifically, in context.
Large Counts Condition
For a proportion procedure, np ≥ 10 and n(1 − p) ≥ 10. Use p₀ for a test, p̂ for an interval.
On FRQs: Getting this wrong (using p̂ in a test) is a very common single-point loss on 1-prop z-test FRQs.
Matched Pairs
A design in which each subject (or naturally-paired subjects) provides two related measurements. Analyzed as a 1-sample t on the differences.
On FRQs: The parameter is μd, the mean difference. Never run a 2-sample test on paired data.
Chi-Square Test for Independence
Tests whether two categorical variables measured on a single sample are associated in the population.
On FRQs: Compare to Homogeneity: independence = one sample, two variables per subject; homogeneity = multiple samples, one variable each.
Chi-Square Test for Homogeneity
Tests whether the distribution of a categorical variable is the same across two or more populations.
On FRQs: Note: chi-square goodness-of-fit was removed from the AP Statistics course in the 2026–27 redesign.
Least-Squares Regression Line (LSRL)
The line ŷ = a + bx that minimizes the sum of squared residuals.
On FRQs: Interpret slope with "on average" and units. State whether the intercept is meaningful in context.
Residual
Observed y minus predicted y. Positive residuals mean the LSRL underpredicted.
On FRQs: A residual plot with random scatter around 0 supports a linear model; a curve suggests linearity is inappropriate.
r² (Coefficient of Determination)
The fraction of variation in y explained by the linear regression on x, between 0 and 1.
On FRQs: Say "about r²·100% of the variation in y is explained by the linear model with x." Missing "of the variation" costs the point.
Extrapolation
Using a regression model to predict outside the observed range of x. Generally unreliable.
On FRQs: When a question asks you to predict at an x-value outside the data, note the extrapolation risk explicitly.
Binomial Distribution
Counts of successes in n independent trials with constant success probability p. BINS: Binary, Independent, Number fixed, Success probability constant.
On FRQs: The geometric distribution was removed in 2026–27 — only binomial remains for count-based problems.
Bluebook
The College Board's digital testing app. Starting May 2027, the AP Statistics exam is administered fully in Bluebook, with a built-in Desmos graphing calculator and provided formula sheets.
On FRQs: You may still bring an approved graphing calculator — many students use both.
Keep going
Say the words graders reward.
Cramapple's FRQ grading flags exact phrasing — 'on average,' 'in context,' 'reject H₀ at α = 0.05' — so you know where a sentence is losing a point.