Build a Complete Regression Response
In “Common Errors Using a Fitted Line,” you checked variable roles, units, and careful prediction wording. Now bring those skills together in a free-response task. A strong response may need to write or identify the regression equation, use it to predict a response, and justify whether that prediction is appropriate.
The calculation is only one part of the answer. A correct numerical prediction can still be incomplete if it omits the equation, reports no units, or gives no reason the prediction is appropriate. In this tutorial, justification means connecting the data’s pattern and the predictor value to whether the fitted line is a sensible tool for the requested prediction.
A Practical Plan for the Free Response
A prompt may give a fitted equation directly, provide computer output, or provide summary statistics from which you must find the slope and intercept. The earlier tutorials “Computing Slope From r and Summary Statistics” and “Reading Regression Computer Output” cover those calculations and how to identify the coefficients. Here, focus on arranging the work so the equation, prediction, and justification each answer the prompt.
For an equation written as \(\hat{y}=a+bx\), keep \(x\) as the predictor and \(y\) as the response. If the equation is not supplied, calculate or read \(a\) and \(b\), then write the equation before using it. Substituting the requested predictor value gives \(\hat{y}\), a predicted response—not an observed response.
Next, check whether the predictor value lies within the range of \(x\)-values used to fit the line. A prediction within that range is interpolation; one outside the range is extrapolation. As covered in “Predicting Within the Data Range,” extrapolation may be unreliable because the pattern observed in the data may not continue beyond the data range. Range alone is not the entire justification: also consider whether the data show a reasonably linear pattern without an unusual feature that undermines the prediction.
These are checks on using a descriptive regression line for prediction. This task is not a significance test or confidence interval, so do not add conditions for an inference procedure that the prompt did not ask you to perform. Justify with the evidence actually supplied, such as a scatterplot or the stated range of the predictor.
Name what the predictor \(x\) and response \(y\) measure, including their units.
Use the supplied coefficients or calculate them as requested, then write \(\hat{y}=a+bx\) with the correct variables.
Substitute the specified predictor value for \(x\), show the arithmetic, and report \(\hat{y}\) in response units.
Use the observed \(x\)-range and the linear pattern described or shown in the prompt. State whether the prediction is interpolation or extrapolation.
Worked Example: Find the Line, Predict, and Justify
Worked Example: Find the Line, Predict, and Justify
A fictional wellness class records weekly exercise time and resting pulse rate for a group of adults. Let \(x\) be exercise time in hours per week and \(y\) be resting pulse rate in beats per minute. The data have \(\bar{x}=4.5\), \(s_x=1.8\) hours per week, \(\bar{y}=71.2\) beats per minute, \(s_y=8.4\) beats per minute, and \(r=-0.72\). The observed exercise times range from 1 to 8 hours per week. A scatterplot shows a reasonably linear negative association with no obvious influential outlier. Write the least-squares regression equation, predict the resting pulse rate for someone who exercises 6 hours per week, and justify the prediction.
State. The predictor is weekly exercise time \(x\), in hours per week. The response is resting pulse rate \(y\), in beats per minute.
Plan. Use \(b=r(s_y/s_x)\) for the slope, then \(a=\bar{y}-b\bar{x}\) for the intercept. Substitute \(x=6\) to find the predicted response. Finally, check the scatterplot description and whether 6 is within the observed predictor range.
Do. First calculate the slope:
The slope is in beats per minute per hour per week. Next calculate the intercept:
Thus, the regression equation is:
Using \(x=6\) hours per week:
Conclude. For an adult who exercises 6 hours per week, the fitted line predicts a resting pulse rate of about 66.2 beats per minute. This is an interpolation because 6 hours per week is within the observed range of 1 to 8 hours per week. The described scatterplot is reasonably linear and has no obvious influential outlier, which supports using the line for this prediction.
Notice how the justification refers to evidence rather than simply declaring the prediction “valid.” The predictor value is within the data range, and the prompt describes a reasonably linear pattern. The conclusion also says “predicts,” names the predictor value, and gives the response units.
Worked Example: Use Coefficients From Output
Worked Example: Use Coefficients From Output
A fictional school garden tracks the number of days after planting and the height of a particular type of seedling. Let \(x\) be days after planting and \(y\) be height in centimeters. Computer output gives an intercept of 8.7 and a slope of 1.25. The recorded planting ages range from 4 to 28 days, and the scatterplot shows a roughly linear upward pattern without a conspicuous outlier. Write the regression equation, predict the height at 20 days, and explain whether using the line is appropriate.
State. Seedling age in days is the predictor, and seedling height in centimeters is the response.
Plan. Use the intercept and slope from the output to write the equation with a hat on the predicted response. Evaluate it at \(x=20\), then compare 20 days with the observed range and use the scatterplot description in the justification.
Do. The regression equation is:
For a seedling age of 20 days:
Conclude. At 20 days after planting, the fitted line predicts a seedling height of 33.7 centimeters. Because 20 days is between the observed ages of 4 and 28 days, this is interpolation. The roughly linear pattern described in the prompt supports using the line for this prediction.
The equation keeps the response in the \(\hat{y}\) position and the predictor beside the slope. That detail matters: a correct coefficient pair written with the variables reversed would describe a different regression. Also, the intercept’s interpretation is not needed to answer this prompt; include it only if asked or if it helps explain the equation.
Worked Example: Calculate It, but Question the Extrapolation
Worked Example: Calculate It, but Question the Extrapolation
A fictional study of a mobile game relates weekly practice time to the number of levels completed. Let \(x\) be practice time in hours per week and \(y\) be levels completed. The fitted line is \(\hat{y}=2.1+0.84x\). The observed practice times range from 8 to 30 hours per week. A prompt asks for a prediction at 45 hours per week and asks whether it is trustworthy.
State. Practice time is the predictor in hours per week, and levels completed is the response in levels.
Plan. Substitute \(x=45\) to calculate the line’s numerical prediction. Then compare 45 with the observed range. The calculation can be made even if the prediction is an extrapolation; the range and available pattern evidence determine how cautiously to interpret it.
Do.
Conclude. The fitted line predicts 39.9 levels for a player practicing 45 hours per week. This prediction is an extrapolation because 45 hours is above the observed maximum of 30 hours per week. It may not be trustworthy: the data do not show whether the linear pattern continues that far beyond the observed range.
Do not confuse “the equation gives a number” with “the prediction is well supported.” The arithmetic is still correct, but the conclusion must make the limitation clear. In an AP response, saying that this is extrapolation and explaining that the relationship may not continue beyond the data range is stronger than merely calling the value “too high.”
What a Strong AP Response Makes Visible
Scoring expectations vary by prompt, so do not assume every question assigns the same number of points to each part. Still, a response is easier to evaluate when it makes each requested task explicit. Treat the following as a practical checklist of common scoring components, not as an official point-by-point rubric.
| Response component | What to show | What often loses credit |
|---|---|---|
| Equation | Correct intercept and slope, with the response predicted by the predictor. | Reversing the variables, copying a coefficient incorrectly, or omitting the equation when requested. |
| Prediction | Substitution of the specified \(x\)-value and a response with appropriate units. | Using the wrong input, giving only an unexplained number, or attaching predictor units to the result. |
| Justification | Reference to the observed \(x\)-range and evidence that the relationship is reasonably linear. | Saying “because the line fits” without evidence, or ignoring extrapolation. |
| Conclusion | A sentence in context identifying what the line predicts and for which predictor value. | Presenting the prediction as an exact observed outcome or as a causal effect. |
If a question gives a scatterplot, describe the pattern you see rather than relying on correlation alone. The earlier tutorial “Writing a Correlation Conclusion on the AP Exam” covers contextual descriptions of \(r\); for a prediction justification, \(r\) may support direction and strength, but it does not by itself show whether a particular prediction is safe. A scatterplot can reveal curvature or an unusual point that a single correlation value does not explain.
Common Mistakes and AP Exam Tips
- Giving the prediction without the requested equation. If the prompt asks for the equation, write it explicitly before substituting. Do not assume that showing only the final predicted value proves which line you used.
- Using the wrong variable as \(x\). State what each variable measures. Put the predictor in the input position, and keep the predicted response marked as \(\hat{y}\).
- Showing no calculation. Include the substitution and arithmetic. This makes the input clear and can help distinguish a setup error from a calculator or rounding error.
- Omitting response units. A predicted response is measured in the units of \(y\). Include those units in the calculation or contextual conclusion.
- Justifying only with the word “linear.” Connect the justification to the evidence supplied. For example, state that the scatterplot is roughly linear and that the requested predictor value lies within the observed range.
- Calling extrapolation impossible. A line can still be evaluated outside the observed range, but the resulting prediction is less well supported. Explain that the pattern may not continue rather than claiming it cannot continue.
- Claiming certainty or causation. Say that the fitted line predicts a response. Do not say an individual will definitely have that response, or that changing the predictor causes the change, unless the study design supports a causal conclusion.
A concise full-credit-style response does not need long explanations. It does need all requested pieces, accurate calculations, and a context-based justification. Keep full precision through the calculation and round the final prediction to a sensible level, following the guidance in “Prediction Rounding and Units.”
Check Your Understanding
For each situation, identify the equation, prediction, or justification that a complete AP-style response should include.
- A line predicting battery life in hours from screen-on time in hours is \(\hat{y}=14.2-0.65x\). Predict battery life for \(x=8\), and write a contextual sentence with units.
- A regression output gives an intercept of 3.4 and a slope of 2.1. Write the generic regression equation using \(x\) and \(y\). Which variable is the predictor?
- The observed \(x\)-values range from 12 to 40. Is a prediction at \(x=25\) interpolation or extrapolation? What other information would help justify using the line?
- A fitted line predicts a response at \(x=55\), but the observed \(x\)-range is 10 to 45. What should the justification say about this prediction?
- Why is “the correlation is strong, so this prediction is definitely accurate” not a complete or careful justification?