A regression line is used to predict future values of y for x-values beyond the range of the observed data. This practice is called extrapolation. Why is extrapolation considered unreliable?
- AThe linear relationship observed within the data range may not hold outside of it, making predictions unreliable.Correct
- BThe regression equation changes when new data are added beyond the original range.
- CExtrapolation requires a larger sample size to produce accurate predictions.
- DThe y-intercept becomes invalid when predicting values beyond the range of x.
Why A is right
Extrapolation is unreliable because the linear model was fit only to data within a certain range of x-values. There is no guarantee that the linear trend continues outside this range — the true relationship may curve, plateau, or change direction. Predictions outside the data range are speculative.
Why the other 3 choices miss
- B
- Incorrect because the regression equation, once fitted, does not change. However, applying it outside the fitted range is the problem, not the equation itself changing.
- C
- Incorrect because sample size affects the precision of the fitted line within the data range but does not resolve the fundamental problem of extrapolation — the unknown behavior of the relationship outside the observed range.
- D
- Incorrect because the y-intercept is simply the predicted value when x = 0. Its interpretation may be meaningful or not, but it does not specifically become invalid due to extrapolation. The issue is applying the entire model outside the data range.
A model predicts weekly sales (y, in hundreds) from advertising spending (x, in thousands of dollars). The regression equation is y-hat = 5 + 2.3x, and r^2 = 0.81. A marketing manager uses this equation to predict sales for an advertising budget of $50,000 (x = 50). The data used to build the model had advertising budgets ranging from $2,000 to $15,000. This prediction is problematic because
- Aadvertising spending cannot be used as an explanatory variable
- Bthe model should use log-transformed data
- Cthe prediction involves extrapolation far beyond the range of the observed dataCorrect
- Dthe r^2 is not high enough for prediction
Why C is right
The model was built using advertising budgets from $2,000 to $15,000 (x = 2 to x = 15 in thousands). Predicting for x = 50 ($50,000) is extreme extrapolation. The linear relationship between advertising and sales may not hold at much higher spending levels — returns on advertising often diminish. The model's predictions outside its data range are unreliable.
Why the other 3 choices miss
- A
- Incorrect. Advertising spending is a perfectly reasonable explanatory variable for sales. The issue is predicting at a value far outside the range of the data used to build the model.
- B
- Incorrect. There is no indication that a log transformation is needed. The issue is extrapolation, not model specification.
- D
- Incorrect. r^2 = 0.81 indicates a strong linear relationship within the observed data range. The problem is not model quality but the prediction being outside the data range.
A regression analysis of the relationship between hours of exercise per week (x) and resting heart rate (y) includes data from adults aged 20-60. The regression equation is y-hat = 80 - 1.5x. A personal trainer uses this equation to predict the resting heart rate of a 75-year-old who exercises 3 hours per week. Which concern is most relevant?
- AThe r^2 value must be very low for a model with a negative slope
- BThe prediction is unreliable because the 75-year-old is outside the population used to develop the model, and age may modify the exercise-heart rate relationshipCorrect
- CThe y-intercept of 80 is unrealistically high
- DThe model underestimates the heart rate because the slope is negative
Why B is right
The model was developed using adults aged 20-60. Applying it to a 75-year-old involves extrapolating to a population not represented in the data. Age is likely a lurking variable that modifies the exercise-heart rate relationship (older adults may respond differently to exercise). The prediction may be unreliable because the model's assumptions may not hold for elderly individuals.
Why the other 3 choices miss
- A
- Incorrect. A negative slope does not imply a low r^2. Strong negative linear relationships can have high r^2 values.
- C
- Incorrect. A resting heart rate of 80 bpm for someone who exercises 0 hours per week is a reasonable baseline value. The y-intercept is not problematic.
- D
- Incorrect. A negative slope is expected (more exercise is associated with lower resting heart rate). The sign of the slope is not the concern; the concern is applying the model outside its intended population.