Translate Output into Meaning
A regression table can give precise numbers without telling a reader what those numbers mean. In “Describing Method, Data, and Variables in a Report,” you identified the individuals and variables in an analysis. Now use that context to write sentences explaining three regression results: the slope, the intercept, and \(r^2\).
The goal is not to repeat a number with a statistic’s name attached. A useful interpretation identifies the variables, gives the direction and units where appropriate, and states what the model supports. Keep the distinction from “Explaining Regression to a Nontechnical Audience” in mind: plain language should make the result easier to understand, not make the claim stronger than the evidence.
Turn the Slope into a Sentence
In a least-squares regression line, \(\hat{y}=a+bx\), the slope is \(b\). It describes how much the model’s predicted response changes when the explanatory variable increases by one unit. The slope’s sign gives the direction of this change: a positive slope means the predicted response increases, and a negative slope means it decreases.
The slope has units formed from the response units per explanatory-variable unit. For example, if \(y\) is water use in liters per day and \(x\) is sunlight in hours per day, the slope is measured in liters per day for each additional hour of sunlight per day. Keep those units with the interpretation; saying only “the response changes by 0.75” leaves the reader unsure what the number measures.
The wording “the model predicts” matters. A slope describes the fitted linear relationship in the data; it does not say that every individual response changes by exactly that amount. Nor does an association alone establish that changing \(x\) causes \(y\) to change. As in “Writing a Conclusion Without Overclaiming,” keep a descriptive interpretation tied to the observed cases and setting.
Worked Example: Interpret a Positive Slope
Situation. In an invented greenhouse project, students recorded sunlight hours per day and daily water use in liters for 28 potted herbs. A regression output table gives a slope of \(0.75\) liters per day for each sunlight hour per day.
| Statistic | Output |
|---|---|
| Slope, \(b\) | 0.75 |
| Explanatory variable, \(x\) | Sunlight hours per day |
| Response variable, \(y\) | Water use in liters per day |
Plan. Identify the direction and units of the slope. The explanatory variable increases by one sunlight hour per day, and the predicted response changes by \(0.75\) liters of water per day.
Do. The slope interpretation is: “For these potted herbs, each additional hour of sunlight per day is associated with a predicted increase of \(0.75\) liters in daily water use, according to the fitted linear model.”
Conclude. The sentence identifies the cases and variables, gives the positive direction, and includes the response units per explanatory-variable unit. It describes an association in the data; it does not claim that additional sunlight causes each herb to use exactly \(0.75\) more liters per day.
Interpret the Intercept Carefully
The intercept is \(a\), the model’s predicted response when \(x=0\). Its units are the same as the response variable’s units. A standard interpretation states what response value the line predicts at zero on the explanatory-variable scale.
That mathematical interpretation does not always produce a useful practical statement. If \(x=0\) is outside the observed range, the interpretation may rely on extrapolation. The line still has an intercept, but the data may not support treating its value as a dependable prediction for cases with \(x=0\). Apply the range-checking ideas from “Writing an Extrapolation Critique,” and distinguish the number’s meaning in the equation from its practical meaning in context.
Worked Example: Interpret an Intercept Outside the Data Range
Situation. An invented analysis uses the number of vendor booths at a community event to predict setup time in minutes. The regression output gives an intercept of \(7.6\) minutes. The events in the data had between 8 and 32 booths.
| Statistic | Output |
|---|---|
| Intercept, \(a\) | 7.6 minutes |
| Observed booth-count range | 8 to 32 booths |
Plan. State what the fitted line predicts at zero booths, then compare zero with the observed range. The intercept is in minutes because setup time is the response.
Do. The mathematical interpretation is: “The fitted line predicts a setup time of \(7.6\) minutes for an event with zero vendor booths.” However, zero booths is below the observed range of 8 to 32 booths.
Conclude. The intercept has a precise role in the equation, but its practical meaning is limited here: the data do not include events with zero booths, so this is an extrapolation. Do not conclude that an event with no booths would actually take \(7.6\) minutes to set up.
Explain \(r^2\) as Variation Accounted For
The coefficient of determination, \(r^2\), summarizes how much of the variation in the observed response values is accounted for by the linear regression model using the explanatory variable. Since output often reports \(r^2\) as a decimal, convert it to a percentage by multiplying by 100. The percentage refers to variation in the response, not to the percentage of individual responses predicted correctly.
A careful interpretation names the response variable, the explanatory variable, and the linear model. For example: “About 81% of the variation in daily solar energy produced among the sampled days is accounted for by the linear regression of daily solar energy on cloud cover.” The percentage has no measurement units.
The remaining percentage is variation not accounted for by this linear model. It does not identify a particular cause, and it does not mean that the model is “wrong” for that percentage of cases. As emphasized in “Effect of Unusual Points on \(r^2\) and \(s\),” \(r^2\) describes variation accounted for, while \(s\) describes typical residual size in response units. These statistics answer different questions.
Worked Example: Interpret \(r^2\) in Context
Situation. An invented environmental data set records daily cloud cover, as a percentage, and solar energy produced, in kilowatt-hours, for a group of monitoring days. A regression output table reports \(r^2=0.81\).
| Statistic | Output |
|---|---|
| Coefficient of determination, \(r^2\) | 0.81 |
| Explanatory variable | Cloud cover, in percent |
| Response variable | Solar energy produced, in kilowatt-hours |
Plan. Convert the decimal to a percentage and describe variation in the response as accounted for by the linear regression using cloud cover. Do not describe the percentage as a success rate or as the amount of energy.
Do. The conversion is \(0.81\times100\%=81\%\). A contextual interpretation is: “About 81% of the variation in solar energy produced among the monitoring days is accounted for by the linear regression model using cloud cover.”
Conclude. The sentence identifies the response, explanatory variable, and model. It does not say that cloud cover causes 81% of solar energy production, that 81% of the days are predicted exactly, or that 81% of the energy is produced by the model.
Put the Three Interpretations Together
When a table includes all three statistics, connect each number to its own meaning rather than blending them into one vague description. The slope gives predicted change and has response-per-explanatory units. The intercept gives a predicted response at \(x=0\) and has response units. \(r^2\) gives a proportion of response variation accounted for by the linear model and has no units.
Worked Example: Write a Complete Interpretation from One Output Table
Situation. An invented study records the number of minutes students spend using a language-learning app each day and their vocabulary quiz scores, in points, for 40 students. The fitted line predicts quiz score from app-use time. The output table reports an intercept of 62 points, a slope of 1.4 points per minute, and \(r^2=0.56\). App-use times in the data range from 5 to 30 minutes per day.
| Statistic | Output |
|---|---|
| Intercept, \(a\) | 62 points |
| Slope, \(b\) | 1.4 points per minute |
| \(r^2\) | 0.56 |
Plan. Write one sentence for each statistic. Include the units for the slope and intercept, convert \(r^2\) to a percentage, and check the observed range before giving a practical interpretation of the intercept.
Do. The slope interpretation is: “For each additional minute of daily app use, the fitted line predicts a \(1.4\)-point increase in vocabulary quiz score, on average, among these students.” The intercept interpretation is: “The fitted line predicts a score of 62 points for a student with zero minutes of daily app use.” Since zero minutes is below the observed range of 5 to 30 minutes, that intercept is an extrapolation and has limited practical meaning for these data. For \(r^2\), \(0.56\times100\%=56\%\), so the interpretation is: “About 56% of the variation in vocabulary quiz scores among these students is accounted for by the linear regression of quiz score on daily app-use time.”
Conclude. Together, these sentences explain the model’s predicted change, its prediction at zero, and the proportion of response variation accounted for. They describe an association for the observed students; they do not establish that app use causes higher scores.
Common Mistakes and AP Exam Tips
- Leaving out the variables or units in a slope interpretation. “The slope is 1.4” does not tell the reader what changes or what the number measures. Full-credit wording identifies a one-unit increase in the explanatory variable and a predicted response change in context with units.
- Describing the slope as a guaranteed change for each individual. The slope describes the fitted linear relationship, not an exact change for every case. Use wording such as “the model predicts” or “is associated with a predicted change.”
- Calling the intercept the response when \(x\) is average. The intercept is the predicted response when \(x=0\), not when \(x\) is at its mean. State the zero-value meaning correctly, then consider whether that value is within the observed range.
- Giving an intercept practical importance without checking the range. If zero is outside the observed explanatory-variable values, note the extrapolation and limit the practical claim. The intercept remains part of the equation even if its real-world meaning is weak.
- Interpreting \(r^2\) as a percentage of cases. “The model predicts 56% of students correctly” is not an interpretation of \(r^2\). State that the specified percentage of variation in the response is accounted for by the linear model using the explanatory variable.
- Making a causal claim from a descriptive regression. A positive slope or large \(r^2\) alone does not show that changing the explanatory variable causes a response change. Keep the conclusion about the association supported by the data and collection method.
Before submitting an interpretation, check that a reader could identify which variables are involved, what each number refers to, and what units apply. For the intercept, also check the observed \(x\)-range. For \(r^2\), use “variation in the response accounted for by the linear model,” not a claim about individual accuracy or causation.
Check Your Understanding
For each question, write a contextual sentence and include the units or limitation requested.
- A regression predicts weekly exercise minutes from days per week with a slope of 18. What does the slope mean, including the units?
- A model predicts package delivery time in hours from route length in kilometers. Its intercept is 0.6. State the mathematical interpretation of the intercept and explain what additional information you would check before judging its practical meaning.
- A regression of monthly household water use on household size has \(r^2=0.49\). Write a contextual interpretation of \(r^2\).
- Why is “49% of households are predicted correctly” not an appropriate interpretation of \(r^2=0.49\)?
- A positive slope is reported for an observational data set. What can you say about the association, and what should you avoid claiming about cause and effect?