ANOVA Foundations Study Pack
Kibin's free study pack on ANOVA Foundations includes a 6-section study guide, 25 quiz questions, 30 flashcards, and 5 open-ended Explain review questions. Sign up free to track your progress toward mastery, plus upload your own notes and recordings to create personalized study packs organized by course.
Last updated May 28, 2026
ANOVA Foundations Study Guide
Break down ANOVA from the ground up — covering between- and within-groups variance, the F-statistic ratio, degrees of freedom, and why a significant result still requires post-hoc tests to pinpoint which group means differ.
Key Takeaways
- •ANOVA (Analysis of Variance) tests whether the means of three or more groups differ significantly by comparing the variance between groups to the variance within groups.
- •The core logic of ANOVA partitions total variability in a dataset into two components: variance explained by group membership (between-groups) and variance due to random error (within-groups).
- •The F-statistic is the ratio of Mean Square Between (MSB) to Mean Square Within (MSW); a large F-value suggests that group differences are unlikely to be due to chance alone.
- •One-way ANOVA requires four key assumptions: independence of observations, approximately normal distributions within each group, homogeneity of variance across groups, and a continuous dependent variable.
- •Rejecting the null hypothesis in ANOVA only indicates that at least one group mean differs — it does not identify which specific groups differ, requiring post-hoc tests for pairwise comparisons.
- •Degrees of freedom govern the shape of the F-distribution used to determine the p-value: between-groups df equals k − 1 (where k is the number of groups) and within-groups df equals N − k (where N is total sample size).
- •The ANOVA summary table organizes Sum of Squares, degrees of freedom, Mean Squares, and the F-statistic into a standard format used to report and interpret results.
The Purpose and Logic of ANOVA
Analysis of Variance addresses a fundamental problem in statistics: when comparing more than two group means, running repeated t-tests inflates the probability of a Type I error, so a single omnibus test is needed instead.
Why Multiple t-Tests Create Problems
- •Each individual hypothesis test carries its own alpha-level risk (commonly 0.05) of falsely rejecting a true null hypothesis.
- •With three groups, three pairwise t-tests would be needed, raising the familywise error rate to roughly 1 − (0.95)³ ≈ 0.14 — nearly three times the intended error rate.
- •ANOVA controls this inflation by testing all group means simultaneously in a single procedure.
The Null and Alternative Hypotheses in ANOVA
- •The null hypothesis states that all population means are equal: μ₁ = μ₂ = … = μk.
- •The alternative hypothesis states that at least one population mean differs from the others — it does not specify which one or how many.
- •This omnibus framing means a significant result requires follow-up analysis to locate which differences exist.
Partitioning Variability: The Heart of the ANOVA Framework
ANOVA works by decomposing the total variability observed in a dataset into distinct sources, allowing a direct comparison of how much variation is attributable to group differences versus random noise.
Total Sum of Squares (SST)
- •SST measures the total variability of all data points around the grand mean — the mean calculated across every observation regardless of group.
- •SST is calculated as the sum of squared deviations of each individual score from the grand mean: SST = Σ(xᵢ − x̄grand)².
Sum of Squares Between Groups (SSB)
- •SSB captures how much the group means deviate from the grand mean, weighted by group size: SSB = Σnⱼ(x̄ⱼ − x̄grand)².
- •A large SSB indicates that groups have substantially different means, which is evidence against the null hypothesis.
Sum of Squares Within Groups (SSW)
- •SSW measures variability inside each group — how much individual scores scatter around their own group mean.
- •SSW reflects random error or individual differences that cannot be explained by group membership: SSW = Σ(xᵢ − x̄ⱼ)².
The Fundamental Partition
- •These three quantities are related by the identity SST = SSB + SSW, meaning total variability is fully accounted for by between-group and within-group components.
- •This partition is what makes ANOVA's comparison of variance sources mathematically coherent.
Mean Squares, the F-Statistic, and Degrees of Freedom
Raw sums of squares cannot be directly compared because they depend on different numbers of data points; dividing by the appropriate degrees of freedom converts them into average variances — the Mean Squares — that can be meaningfully ratioed.
Degrees of Freedom in One-Way ANOVA
- •Between-groups degrees of freedom equal k − 1, where k is the number of groups being compared.
- •Within-groups degrees of freedom equal N − k, where N is the total number of observations across all groups.
- •Total degrees of freedom equal N − 1, and dfBetween + dfWithin = dfTotal, mirroring the sum of squares partition.
Mean Square Between (MSB) and Mean Square Within (MSW)
- •MSB = SSB / (k − 1); it estimates the variance among group means, inflated by any real treatment effect.
- •MSW = SSW / (N − k); it estimates variance due to individual differences and measurement error, unaffected by any treatment effect.
- •MSW is also called the error mean square and serves as the baseline estimate of natural variability.
The F-Statistic
- •The F-statistic is the ratio F = MSB / MSW.
- •When the null hypothesis is true, both MSB and MSW estimate the same population variance, so F hovers near 1.
- •When group means truly differ, MSB grows larger than MSW, pushing F above 1 and eventually into the rejection region of the F-distribution.
- •The p-value is determined by comparing the calculated F to a theoretical F-distribution with (k − 1) numerator degrees of freedom and (N − k) denominator degrees of freedom.
Unlock the rest of this study guide
- Access the full study pack
- Track your mastery and be test-day ready
- Upload your own notes to build personalized study guides, quizzes, flashcards, and more
About this Study Pack
Created by Kibin to help students review key concepts, prepare for exams, and study more effectively. This Study Pack was checked for accuracy and curriculum alignment using authoritative educational sources. See sources below.
Sources
Question 1 of 25
Your progress is saved after each question and counts toward mastery.
When comparing three groups using individual t-tests, what is the approximate familywise error rate if each test uses α = 0.05?
Card 1 of 30
Your progress is saved after each card and counts toward mastery.
Concept 1 of 5
Your progress is saved after each concept and counts toward mastery.
The Purpose of ANOVA and Familywise Error
Explain why researchers use ANOVA instead of running multiple t-tests when comparing three or more groups. What is the familywise error rate, and how does ANOVA address the problem it creates?
More in Statistics
See all topics →Binomial Distributions
Master the binomial distribution from the ground up — covering the four conditions for a binomial setting, the probability formula P(X = k) = C(n, k) · pᵏ · (1 - p)ⁿ⁻ᵏ, and how to find mean and standard deviation using np and √(np(1 - p)).
Central Limit Theorem
Unpack the Central Limit Theorem and see why sample means form a normal distribution as n grows — even when the population isn't normal. Master key mechanics like standard error (σ/√n), the n ≥ 30 rule, and applying z-scores to sample mean problems.
Confidence Level and Margin of Error
Unpack the mechanics of confidence intervals and margin of error, from how critical values (z* and t*) are chosen to how sample size affects interval width. Master the distinction between z- and t-distributions and what confidence level actually means across repeated samples.
Data Visualization and Distribution Shapes
Visualize how raw data takes shape through histograms, dot plots, and box plots while mastering symmetric, skewed, and uniform distributions — and learn why skewness shifts the mean toward the tail but leaves the median largely unaffected.
Experimental Design and Bias
Unpack the core principles of experimental design — from random assignment and confounding variables to single- and double-blind procedures, placebo controls, and bias types — so you can confidently distinguish true experiments from flawed ones.
Hypothesis Testing Logic
Unpack the logic behind hypothesis testing, from null and alternative hypotheses to p-values, significance levels, and Type I and II errors. This pack clarifies how test statistics and tail direction shape your reject-or-fail-to-reject decision.
Measures of the Center of the Data
Master the three measures of center — mean, median, and mode — and learn how each responds to outliers, skewness, and weighted values. Understand when to use the median over the mean for skewed data like income, and how symmetric vs. skewed distributions shift these measures apart.
Measures of Variability
Unpack the core measures of spread — range, standard deviation, variance, and IQR — and learn when to use each based on data shape and outliers. Covers Chebyshev's theorem and the n − 1 correction for unbiased sample estimates.
Percentiles and Z-Scores
Unpack the relationship between percentiles, z-scores, and the standard normal distribution — covering the z = (x − μ) / σ formula, IQR-based outlier detection, and how standardized scores translate raw data into comparable percentile ranks.
Prediction
Master the mechanics of simple linear regression, from calculating slope and y-intercept using ŷ = a + bx to interpreting r² and avoiding extrapolation pitfalls. Covers when correlation is significant enough to predict and why the point (x̄, ȳ) always anchors the line.