Unit 5 · Regression Analysis
Unit 5 · Regression Analysis
- 10–20% of the multiple-choice section
- 5 original figures
- clean-room review
This guide organizes Regression Analysis around one repeatable exam decision: describe linear association, interpret slope and residuals in context, and separate predictive fit from causal explanation. In Regression Analysis, formulas and vocabulary belong to an evidence chain rather than an isolated recall list.
- Decision: describe linear association, interpret slope and residuals in context, and separate predictive fit from causal explanation.
- Representation: move deliberately among scatterplot with form, direction, strength, and unusual features, residual plot, least-squares line with slope triangle.
- Regression Analysis response standard: separate evidence from scope: random selection supports population generalization, random assignment supports causation, and neither can be silently substituted for the other.
What Regression Analysis covers
The frozen taxonomy groups Regression Analysis into 4 exam-facing skill routes. Each Regression Analysis route keeps official topic ownership inside this unit.
Where Regression Analysis sits on the exam
College Board assigns Regression Analysis 10–20% of AP Statistics multiple-choice content. This range is not a share of the total exam score and does not imply a fixed question count or an FRQ allocation.
A statistics-capable graphing calculator and official reference information support computation, not procedure selection or interpretation. Calculator details should always be checked against the current official policy at College Board.
The decision that organizes Regression Analysis
Start with the claim, not the formula
In Regression Analysis, the decisive question is whether you can describe linear association, interpret slope and residuals in context, and separate predictive fit from causal explanation. The prompt may look computational, but scatterplot with form, direction, strength, and unusual features must agree with the relationship 'A residual equals observed response minus predicted response.' before the result is defensible. Begin by trying to inspect the scatterplot and residual pattern before interpreting correlation or using the fitted line. That move keeps residual plot paired with its stated conditions and heads off the neighboring error of using correlation with categorical variables.
Build an evidence chain
The Regression Analysis evidence chain begins with the situation 'A least-squares model predicts exam score from study hours, and a student with 6 hours scores 4 points above the prediction.' and moves through scatterplot with form, direction, strength, and unusual features, residual plot, or least-squares line with slope triangle. Each Regression Analysis surface should lead to one named relationship and one conclusion whose scope is visible. On scatterplot with form, direction, strength, and unusual features, label the measured feature and direction. When the same information is recast as residual plot, preserve the reference point, units, and controlled conditions. Use least-squares line with slope triangle as the final consistency check rather than leaving the answer as calculator output.
Three relationships worth being able to explain
A residual equals observed response minus predicted response. For Regression Analysis, test this statement against scatterplot with form, direction, strength, and unusual features and explicitly name which quantity changes. When those Regression Analysis conditions are absent, give a conditional prediction instead of a numerical claim.
The least-squares line minimizes the sum of squared vertical residuals. Use this Regression Analysis connection to reconcile residual plot with least-squares line with slope triangle. A Regression Analysis disagreement points to a sign, denominator, reference, or model error that must be diagnosed before the response is finalized.
The coefficient of determination describes the fraction of response variation explained by the linear model. This relationship marks the boundary next to 'claiming that strong association proves causation.' State the extra condition or observation that the stronger claim would require, especially when the prompt supplies only one representation.
Decision route.
Decision route. For Regression Analysis, follow the evidence in order so a skipped representation or boundary does not create an overclaim.
Read the surface before you solve Regression Analysis
What the representation can tell you
For Regression Analysis, first name whether the prompt gives scatterplot with form, direction, strength, and unusual features, residual plot, or least-squares line with slope triangle. On that Regression Analysis surface, mark axes, labels, units, direction convention, and the relevant population, system, function, market, or chemical process. Describe one visible feature, then connect it to 'The least-squares line minimizes the sum of squared vertical residuals..' Keeping that Regression Analysis observation separate from its explanation makes the inference auditable and exposes any assumption that the picture itself does not show.
Error boundaries that preserve credit
The error boundary for Regression Analysis starts with 'using correlation with categorical variables': return to scatterplot with form, direction, strength, and unusual features and restore the label or condition the shortcut erased. If a solution starts extrapolating far beyond the observed explanatory range, make the intermediate quantity visible on residual plot instead of carrying the step mentally. The remaining boundary is claiming that strong association proves causation. Close a Regression Analysis response by stating what least-squares line with slope triangle establishes and what additional evidence the stronger neighboring claim would need.
Representation lab.
Representation lab. This Regression Analysis drawing is a clean-room schematic, not official exam data; read its axes and labels before importing a memorized rule.
Scatterplots and Quantitative Association
Recognize and route the skill
Scatterplots and Quantitative Association is a decision cluster inside Regression Analysis; cues include scatterplot, explanatory-axis, response-axis, direction-form-strength. For Scatterplots and Quantitative Association, state the target claim in words and route it through the unit decision: describe linear association, interpret slope and residuals in context, and separate predictive fit from causal explanation. Routing Scatterplots and Quantitative Association through that decision prevents a familiar operation from answering a neighboring question.
Operate, check, and communicate
For Scatterplots and Quantitative Association, check scatterplot with form, direction, strength, and unusual features, then apply this relationship only when its conditions match: A residual equals observed response minus predicted response. Keep the Scatterplots and Quantitative Association labels, sign, and context attached to the result. The adjacent Scatterplots and Quantitative Association error is using correlation with categorical variables. To repair Scatterplots and Quantitative Association, restore the missing condition, restart from inspect the scatterplot and residual pattern before interpreting correlation or using the fitted line, and finish with evidence, consequence, and a bounded contextual claim.
Correlation
Recognize and route the skill
Correlation is a decision cluster inside Regression Analysis; cues include correlation-coefficient, linear-association, unitless, nonresistant. For Correlation, state the target claim in words and route it through the unit decision: describe linear association, interpret slope and residuals in context, and separate predictive fit from causal explanation. Routing Correlation through that decision prevents a familiar operation from answering a neighboring question.
Operate, check, and communicate
For Correlation, check residual plot, then apply this relationship only when its conditions match: The least-squares line minimizes the sum of squared vertical residuals. Keep the Correlation labels, sign, and context attached to the result. The adjacent Correlation error is extrapolating far beyond the observed explanatory range. To repair Correlation, restore the missing condition, restart from inspect the scatterplot and residual pattern before interpreting correlation or using the fitted line, and finish with evidence, consequence, and a bounded contextual claim.
Linear Prediction and Residual Analysis
Recognize and route the skill
Linear Prediction and Residual Analysis is a decision cluster inside Regression Analysis; cues include predicted-value, residual, residual-plot, extrapolation. For Linear Prediction and Residual Analysis, state the target claim in words and route it through the unit decision: describe linear association, interpret slope and residuals in context, and separate predictive fit from causal explanation. Routing Linear Prediction and Residual Analysis through that decision prevents a familiar operation from answering a neighboring question.
Operate, check, and communicate
For Linear Prediction and Residual Analysis, check least-squares line with slope triangle, then apply this relationship only when its conditions match: The coefficient of determination describes the fraction of response variation explained by the linear model. Keep the Linear Prediction and Residual Analysis labels, sign, and context attached to the result. The adjacent Linear Prediction and Residual Analysis error is claiming that strong association proves causation. To repair Linear Prediction and Residual Analysis, restore the missing condition, restart from inspect the scatterplot and residual pattern before interpreting correlation or using the fitted line, and finish with evidence, consequence, and a bounded contextual claim.
Least-Squares Regression
Recognize and route the skill
Least-Squares Regression is a decision cluster inside Regression Analysis; cues include least-squares-line, slope-in-context, intercept-in-context, coefficient-of-determination. For Least-Squares Regression, state the target claim in words and route it through the unit decision: describe linear association, interpret slope and residuals in context, and separate predictive fit from causal explanation. Routing Least-Squares Regression through that decision prevents a familiar operation from answering a neighboring question.
Operate, check, and communicate
For Least-Squares Regression, check scatterplot with form, direction, strength, and unusual features, then apply this relationship only when its conditions match: A residual equals observed response minus predicted response. Keep the Least-Squares Regression labels, sign, and context attached to the result. The adjacent Least-Squares Regression error is using correlation with categorical variables. To repair Least-Squares Regression, restore the missing condition, restart from inspect the scatterplot and residual pattern before interpreting correlation or using the fitted line, and finish with evidence, consequence, and a bounded contextual claim.
How the AP Statistics assesses Regression Analysis
Unit ranges describe the multiple-choice section only. Free-response work can combine content across units, so no per-unit FRQ share is inferred.
| Item | Weight / count | What it means |
|---|---|---|
| Multiple choice | 42 questions · 90 minutes · 50% | Single-select questions appear in Bluebook; current planning supports both discrete and stimulus-linked reasoning without promising an unverified set count. |
| Free response | 4 questions · 90 minutes · 50% | Responses are typed in Bluebook and include multi-focus and inference work. |
| Calculator | Statistics-capable graphing calculator | A graphing calculator can execute arithmetic, but the response must still identify conditions, parameters, and a contextual conclusion. |
| Unit weight | 10–20% of the multiple-choice section | This published range applies to multiple choice, not to a promised count or an FRQ allocation. |
| Response evidence | Represent · relate · verify | Separate evidence from scope: random selection supports population generalization, random assignment supports causation, and neither can be silently substituted for the other. |
Choose the first defensible move in Regression Analysis
Read each question as it appears on the exam paper, choose your answer, then tap it to see the feedback.
Practice 1: Interpret a positive residual
A least-squares model predicts exam score from study hours. A student who studies for 6 hours scores 4 points above the model’s prediction.
Which first step best preserves a valid regression analysis?
Choose the best answer.
A. Inspect the scatterplot and residual pattern before interpreting correlation or using the fitted line.
✅ Correct! 😊 The plots reveal whether a linear model is appropriate and whether unusual features matter. The residual can then be interpreted as observed minus predicted in score points.
B. Use correlation even if study time is converted to categories.
❌ Not quite 😭 Correlation describes association between two quantitative variables. Converting study time to categories changes the information and makes correlation inappropriate for that version of the variable.
Prep with Sia Agent
Free download for Mac and Windows. Choose your computer:
C. Use the line to predict scores far beyond the observed range of study hours.
❌ Not quite 😭 A fitted relationship may not continue outside the observed explanatory range. Such extrapolation can produce unsupported predictions even when the line fits the recorded data.
Prep with Sia Agent
Free download for Mac and Windows. Choose your computer:
D. Treat the residual of 4 as proof that studying caused the score increase.
❌ Not quite 😭 A residual only compares an observed response with its prediction. It does not establish causation or measure how much studying changed the score.
Prep with Sia Agent
Free download for Mac and Windows. Choose your computer:
Full explanation
- Step 1Inspect the scatterplot for form, direction, strength, and unusual observations.
- Step 2Inspect residuals for systematic patterns that would challenge a linear model.
- Step 3Use residual = observed score − predicted score, so a residual of 4 means four points above prediction.
- Step 4Keep prediction and association separate from any causal claim.
Practice 2: Interpret slope in context
For a sample of delivery trips, the least-squares line predicting delivery time y, in minutes, from route distance x, in miles, is ŷ = 12.5 + 2.8x.
Interpret the slope and explain whether it supports a causal conclusion about increasing route distance.
Show your work.
Answer in context.
A. Every additional mile causes every delivery to take exactly 2.8 more minutes.
❌ Not quite 😭 The slope describes a predicted average change, not an exact change for every trip. A regression from observational data also does not by itself establish causation.
Prep with Sia Agent
Free download for Mac and Windows. Choose your computer:
B. For trips represented by the data, each additional mile is associated with a predicted delivery time 2.8 minutes higher, without proving causation.
✅ Correct! 😊 This gives the slope’s units, direction, and predictive meaning in context. It also keeps an observed linear association separate from a causal claim.
C. A trip of zero miles must take exactly 12.5 minutes, so the slope is reliable at all distances.
❌ Not quite 😭 The intercept is a model prediction and may lack practical meaning if zero miles is outside the observed range. It does not validate the slope at unobserved distances.
Prep with Sia Agent
Free download for Mac and Windows. Choose your computer:
D. Delivery time increases by 12.5 minutes whenever distance increases by 2.8 miles.
❌ Not quite 😭 This swaps the roles of intercept and slope. The coefficient 2.8 represents predicted minutes per additional mile; 12.5 is the predicted intercept.
Prep with Sia Agent
Free download for Mac and Windows. Choose your computer:
Full explanation
- Step 1Identify 2.8 as the slope, with units of minutes per mile.
- Step 2Interpret it as a 2.8-minute increase in predicted delivery time for each additional mile.
- Step 3Limit the interpretation to the range and setting represented by the data.
- Step 4Do not claim causation without a design that supports causal inference.
Working language for Regression Analysis
- Scatterplots and Quantitative Association
- In Regression Analysis, Scatterplots and Quantitative Association names the linked decisions for recognizing the evidence, selecting a valid relationship, and stating a contextual conclusion.
- Correlation
- In Regression Analysis, Correlation names the linked decisions for recognizing the evidence, selecting a valid relationship, and stating a contextual conclusion.
- Linear Prediction and Residual Analysis
- In Regression Analysis, Linear Prediction and Residual Analysis names the linked decisions for recognizing the evidence, selecting a valid relationship, and stating a contextual conclusion.
- Least-Squares Regression
- In Regression Analysis, Least-Squares Regression names the linked decisions for recognizing the evidence, selecting a valid relationship, and stating a contextual conclusion.
- Regression Analysis
- The official Regression Analysis frame that connects its frozen skill leaves through one evidence-preserving decision route for AP Statistics.
- evidence chain
- The Regression Analysis sequence from observation to representation, relationship, operation, verification, and a claim limited by the available evidence.
- representation check
- A deliberate inspection of labels, axes, units, direction, population, system, or market before solving a Regression Analysis problem.
- error boundary
- A condition that separates a warranted Regression Analysis inference from a stronger neighboring claim that the prompt does not establish.
Regression Analysis questions students actually ask
What is the first decision in Regression Analysis?
Begin Regression Analysis by deciding how to describe linear association, interpret slope and residuals in context, and separate predictive fit from causal explanation. Then inspect the scatterplot and residual pattern before interpreting correlation or using the fitted line. This keeps the Regression Analysis target claim, given conditions, and representation aligned before arithmetic or symbolic manipulation begins.
Which representation should I draw for Regression Analysis?
For Regression Analysis, choose among scatterplot with form, direction, strength, and unusual features, residual plot, least-squares line with slope triangle according to the evidence. Label the Regression Analysis axes, units, system or population, and direction before using the drawing to justify a relationship or numerical result.
How do I repair the most common Regression Analysis shortcut?
In Regression Analysis, watch for using correlation with categorical variables. Return to the Regression Analysis prompt, restore the skipped condition or representation, and rebuild the evidence chain from inspect the scatterplot and residual pattern before interpreting correlation or using the fitted line rather than patching the final line.
What makes a Regression Analysis explanation complete?
In Regression Analysis, a complete explanation names the governing relationship, points to the relevant evidence, states the directional or numerical consequence, and finishes in context. For Regression Analysis, you should separate evidence from scope: random selection supports population generalization, random assignment supports causation, and neither can be silently substituted for the other.
Should I memorize every formula in Regression Analysis?
For Regression Analysis, memorize only what the official reference policy requires, but practice selecting and explaining every relationship. For Regression Analysis, a statistics-capable graphing calculator and official reference information support computation, not procedure selection or interpretation. A Regression Analysis formula is useful only after its variables and assumptions match the prompt.
Continue through all AP Statistics units
A durable study loop for Regression Analysis
Build a one-page decision map for Regression Analysis. Put the question 'describe linear association, interpret slope and residuals in context, and separate predictive fit from causal explanation?' at the center, connect it to scatterplot with form, direction, strength, and unusual features, residual plot, least-squares line with slope triangle, and write the condition that licenses each relationship beside its arrow.
Practice Regression Analysis representation translation in pairs. Convert scatterplot with form, direction, strength, and unusual features into residual plot, then reverse the translation without looking. Any Regression Analysis feature that disappears in one direction identifies a label, unit, or assumption that needs deliberate rehearsal.
Keep a Regression Analysis error log organized by broken step instead of by problem number. When you catch using correlation with categorical variables, record the missing cue and the repair action. Re-solve the Regression Analysis prompt after two days and one week using only that cue.
For timed Regression Analysis work, spend the opening seconds framing the object and expected direction. Then solve the Regression Analysis prompt, verify with a second representation or limiting case, and write the contextual conclusion. This Regression Analysis routine is faster than repairing an answer built on the wrong model.
