- A educational researcher examines the relationship between class size (x) and average test scores (y) across 45 elementary schools. The regression analysis shows y=87.2−1.3x with r2=0.52. The residual plot reveals that schools with class sizes above 28 students show much more scattered residuals than schools with smaller classes. What does this pattern suggest about the model's applicability?
Answer: The increased scatter in residuals for large classes suggests the linear relationship breaks down at higher class sizes, making predictions less reliable for schools with more than 28 students per class.
When analyzing regression models, you need to examine both the overall fit statistics and the residual patterns to assess model reliability. While r2=0.52 tells you that 52% of the variation in test scores is explained by class size overall, the residual plot reveals crucial information about where the model works well and where it breaks down.
The correct answer is D because increasing scatter in residuals at larger class sizes indicates heteroscedasticity - the model's prediction errors become more variable and less reliable in that range. This suggests the linear relationship y=87.2−1.3x may not adequately capture the true relationship for classes above 28 students, making predictions less trustworthy in that region.
Option A incorrectly assumes that overall r2 guarantees uniform reliability across all data ranges. The r2 value represents average explanatory power but doesn't tell you about local breakdowns in the relationship.
Option B misinterprets scattered residuals as evidence that class size doesn't matter. The negative slope and significant r2 clearly show a relationship exists; the scattered residuals just indicate where that relationship becomes less predictable.
Option C makes an illogical leap, suggesting scattered residuals mean the relationship reverses direction. Increased scatter means more uncertainty, not a sign flip in the relationship.
Study tip: Always examine residual plots alongside summary statistics. Patterns in residuals (like increasing scatter, curves, or outlier clusters) reveal where your model's assumptions break down, even when overall fit statistics look decent. - A business analyst examines the relationship between weekly hours worked (x) and productivity score (y) for remote employees. The regression analysis yields y=45.2+2.1x with r=0.43. Management wants to use this model to determine optimal work hours. What is the most significant limitation of this approach?
Answer: The linear model assumes a constant rate of productivity increase per hour, but in reality, productivity may plateau or decrease due to fatigue and burnout at higher hour ranges.
The most significant limitation is that linear models assume a constant rate of change, but productivity typically follows a non-linear pattern where additional hours may yield diminishing returns or even negative effects due to fatigue and burnout. This makes the linear model inappropriate for finding 'optimal' work hours. Choice A incorrectly suggests the model predicts unlimited improvement. Choice B focuses on r² (which would be 0.43² ≈ 0.18) but this isn't the most critical limitation for optimization. Choice D misinterprets the intercept, which represents baseline productivity from factors other than hours worked.
- A researcher collected data on the relationship between years of experience (x) and annual salary in thousands of dollars (y) for software engineers. The linear regression equation is y=65.2+4.8x with r2=0.72. Based on this model, if a software engineer has 8 years of experience, what is the predicted annual salary, and how should this prediction be interpreted?
Answer: The predicted salary is $103,600, but since only 72% of salary variation is explained by experience, there's significant uncertainty and the actual salary could vary considerably.
Using the regression equation: y = 65.2 + 4.8(8) = 65.2 + 38.4 = 103.6 thousand dollars, or $103,600. The r² = 0.72 indicates that 72% of salary variation is explained by experience, meaning 28% is unexplained, creating uncertainty in individual predictions. Choice A incorrectly treats r² as confidence in exact predictions. Choice C calculates the salary incorrectly and misinterprets r² as margin of error. Choice D also calculates incorrectly and misinterprets the prediction as only partial salary.
- A real estate analyst develops a linear regression model to predict house prices based on square footage. The model is y=85,000+120x, where y is price in dollars and x is square footage. The data range from 800 to 3,500 square feet with r2=0.71. A client asks for a price estimate on a 5,200 square foot mansion. How should the analyst respond?
Answer: The predicted price is $709,000, but this estimate is highly unreliable because it requires extrapolation far beyond the observed data range, where the linear relationship may not hold.
When you encounter linear regression problems involving predictions, always consider whether the prediction falls within the original data range. This question tests your understanding of extrapolation versus interpolation.
Let's calculate the predicted price: y=85,000+120(5,200)=85,000+624,000=709,000. So the mathematical prediction is $709,000.
However, the key issue is that 5,200 square feet falls far outside the observed data range of 800-3,500 square feet. This requires extrapolation, which assumes the linear relationship continues unchanged beyond the data. In reality, housing markets often behave differently at extreme values—luxury mansions may have different pricing dynamics than typical homes.
Answer A correctly identifies both the calculated price and the reliability concern due to extrapolation beyond the data range.
Answer B makes a critical error by assuming the r2=0.71 guarantees accuracy across all ranges. The coefficient of determination only describes how well the model fits within the observed data range, not beyond it.
Answer C incorrectly states that regression equations are "undefined" outside the original range. While you can always calculate a prediction mathematically, the issue is reliability, not mathematical impossibility.
Answer D misinterprets what r2 means. The 0.71 value indicates 71% of variation is explained by the model within the data range, but you don't adjust predictions by subtracting the unexplained variation percentage.
Remember: Always check whether predictions require interpolation (within the data range) or extrapolation (beyond it). Extrapolation dramatically reduces reliability, regardless of how strong the original model appears. - A technology company analyzes the relationship between employee age (x) and monthly app downloads they generate (y). The regression yields y=2,450−18x with r2=0.29 and standard error of 420 downloads. Ages in the dataset range from 22 to 58 years. How should management interpret these results for workforce planning decisions?
Answer: The relationship between age and downloads is weak and highly variable, with substantial individual differences that make age a poor predictor for individual performance evaluation or hiring decisions.
When interpreting regression analysis results, you need to carefully consider both the strength of the relationship and the practical implications for decision-making. The key metrics here are the r2=0.29 and the large standard error relative to the predicted values.
The correct answer is D because this model shows a weak predictive relationship. An r2 of 0.29 means only 29% of the variation in downloads is explained by age, leaving 71% unexplained. The standard error of 420 downloads is quite large compared to the range of predicted values (the model predicts roughly 1,054 to 2,054 downloads for ages 58 to 22). This creates wide confidence intervals that make individual predictions unreliable.
Option A incorrectly treats correlation as definitive causation and ignores the weak relationship strength. The negative slope exists but doesn't justify such extreme hiring policies given the poor predictive power.
Option B misinterprets what r2=0.29 means—it indicates a weak relationship, not a strong enough one to justify age-based quotas. This would likely be discriminatory and ineffective.
Option C overstates the precision of the model. The large standard error means individual predictions are highly uncertain, making this unsuitable for evaluating specific employees.
Study tip: When analyzing regression results, always evaluate both statistical significance and practical significance. A low r2 (typically below 0.5) combined with high standard errors suggests the model isn't reliable enough for individual-level decisions, even if a relationship technically exists. - A sports analyst uses linear regression to model the relationship between a basketball player's practice hours per week (x) and free throw percentage (y). The model is y=42.8+3.2x with r2=0.38. The data includes players practicing 5-25 hours per week. What is the most critical consideration when applying this model?
Answer: The model suggests a positive relationship between practice and performance, but individual talent, technique, and other factors account for most of the variation in free throw percentage.
With r² = 0.38, only 38% of the variation in free throw percentage is explained by practice hours, meaning 62% is due to other factors like natural talent, technique, mental state, etc. The model shows practice has a positive effect, but individual predictions will have substantial uncertainty. Choice A ignores the upper limit of free throw percentage (100%) and model limitations. Choice B overstates the impact of moderate r² - the model still shows meaningful relationships. Choice D reverses the interpretation of r² and incorrectly claims causation.
- A marketing analyst uses linear regression to model the relationship between advertising spending (in thousands of dollars) and monthly sales revenue (in thousands of dollars). The regression output shows: y=12.5+3.2x, where x represents advertising spending and y represents sales revenue. The correlation coefficient is r=0.78 and the coefficient of determination is r2=0.61. What is the most accurate interpretation of the coefficient of determination in this context?
Answer: Approximately 61% of the variation in monthly sales revenue can be explained by the linear relationship with advertising spending, while 39% is due to other factors not included in the model.
The coefficient of determination (r²) represents the proportion of variance in the dependent variable (sales revenue) that is explained by the independent variable (advertising spending) through the linear relationship. An r² of 0.61 means 61% of the variation in sales revenue can be attributed to its linear relationship with advertising spending. Choice B incorrectly interprets r² as probability of causation. Choice C confuses r² with prediction accuracy or confidence intervals. Choice D misinterprets r² as the proportion of total revenue rather than explained variance.
- A health researcher analyzes the relationship between daily exercise minutes (x) and resting heart rate in beats per minute (y) for a sample of adults. The regression output shows y=78.5−0.25x with r2=0.56. If the standard error of the estimate is 8.2 bpm, what can be concluded about the practical usefulness of this model for individual health assessments?
Answer: The model shows a meaningful relationship between exercise and heart rate, but individual predictions have substantial uncertainty, with typical errors around ±8.2 bpm from predicted values.
While the model shows a meaningful relationship (r² = 0.56 explains substantial variation), the standard error of 8.2 bpm indicates considerable uncertainty in individual predictions. For health assessments, this level of uncertainty (±8.2 bpm) is quite substantial relative to typical heart rate ranges. Choice A overstates precision and misinterprets the slope as exact individual change. Choice C incorrectly compares standard error to slope coefficient - they measure different things. Choice D completely misinterprets what r² represents.
