Math 2 Quiz: Evaluating Linear Model Fit
15 questions · exam conditions
0:00
Evaluating Linear Model FitQuestion 1 of 15

An economist studies the relationship between years of education and income using linear regression on a sample of 200 adults. The analysis yields R2=0.71R^2 = 0.71 and normally distributed residuals with no apparent pattern. However, when the economist separates the data by age groups, the slope of the education-income relationship appears steeper for younger adults than older adults. What does this suggest about the linear model?

The linear model is appropriate for the overall population, but separate models should be considered for different age groups to improve precision.
The linear model is inappropriate because different slopes across age groups indicate that age moderates the relationship, violating linearity assumptions.
The linear model is appropriate because high R2R^2 and good residual properties indicate strong overall fit regardless of subgroup differences.
The linear model is inappropriate because varying slopes indicate measurement error or data collection problems that invalidate the analysis.
← Back to quizzes

Math 2 Quiz

Math 2 Quiz: Evaluating Linear Model Fit

Practice Evaluating Linear Model Fit in Math 2 with focused quiz questions that help you check what you know, review explanations, and build confidence with test-style prompts.

What this quiz covers

This quiz focuses on Evaluating Linear Model Fit, giving you a quick way to practice the rules, question types, and explanations that matter most for Math 2.

How to use this quiz

Try each quiz question before looking at the correct answer. Use the explanations to review missed ideas, then come back to similar questions until the pattern feels familiar.

All questions

Question 1

An economist studies the relationship between years of education and income using linear regression on a sample of 200 adults. The analysis yields R2=0.71R^2 = 0.71 and normally distributed residuals with no apparent pattern. However, when the economist separates the data by age groups, the slope of the education-income relationship appears steeper for younger adults than older adults. What does this suggest about the linear model?

  1. The linear model is appropriate for the overall population, but separate models should be considered for different age groups to improve precision. (correct answer)
  2. The linear model is inappropriate because different slopes across age groups indicate that age moderates the relationship, violating linearity assumptions.
  3. The linear model is appropriate because high R2R^2 and good residual properties indicate strong overall fit regardless of subgroup differences.
  4. The linear model is inappropriate because varying slopes indicate measurement error or data collection problems that invalidate the analysis.
Explanation: When you encounter regression problems involving subgroup differences, you're dealing with questions about model specification and whether additional variables or interactions should be included. The key insight here is distinguishing between a model that works well overall versus one that could be improved by accounting for subgroup differences. The original model shows strong performance with R2=0.71R^2 = 0.71 and well-behaved residuals, indicating it captures the general education-income relationship effectively. However, the different slopes across age groups suggest that age acts as a moderator variable - the strength of the education-income relationship depends on age. Option A is correct because it recognizes both realities: the overall model is statistically sound, but separate age-specific models would likely provide better precision by capturing the varying relationship strengths. This is a common scenario in applied statistics where aggregate models work but stratified analyses reveal meaningful heterogeneity. Option B incorrectly claims the model violates linearity assumptions. Different slopes across subgroups don't violate linearity - they suggest an interaction effect that could be modeled with age as a moderator variable. Option C makes the mistake of assuming high R2R^2 means no improvement is possible, ignoring the valuable information about age-related differences. Option D incorrectly attributes slope differences to data problems rather than recognizing this as potentially meaningful variation that could inform better modeling. Remember: High overall model fit doesn't preclude the possibility of meaningful subgroup differences that could improve prediction accuracy through stratification or interaction terms.

Question 2

A student analyzes the relationship between temperature and ice cream sales using linear regression. The correlation coefficient is r=0.82r = 0.82 and R2=0.67R^2 = 0.67. When examining the residual plot, the student observes that residuals are larger in magnitude during summer months (higher temperatures) compared to winter months. What is the primary concern with this linear model?

  1. The correlation coefficient of 0.82 is too low to justify using linear regression, as values below 0.90 indicate insufficient linear relationship.
  2. The R2R^2 value of 0.67 means the model explains less than 70% of variance, making it inadequate for practical prediction purposes.
  3. The varying residual magnitudes across temperature ranges violate the constant variance assumption, potentially affecting inference and prediction intervals. (correct answer)
  4. The seasonal pattern in residuals indicates that time should be included as an additional explanatory variable to improve model accuracy.
Explanation: The changing magnitude of residuals across temperature ranges indicates heteroscedasticity (non-constant variance), which violates a key assumption of linear regression and affects the reliability of confidence intervals and significance tests. Choice A sets an arbitrary and too-high threshold for correlation. Choice B incorrectly suggests R² = 0.67 is inadequate when it's actually quite good. Choice D misidentifies the issue as needing additional variables rather than addressing the variance assumption violation.

Question 3

A quality control engineer models the relationship between machine speed and defect rate using linear regression. The residual analysis shows: (1) residuals are approximately normally distributed, (2) no obvious pattern in residuals versus fitted values, and (3) one data point has a residual more than 3 standard deviations from zero. The engineer notes this extreme point corresponds to a day when the machine was being serviced. How should this affect the model evaluation?

  1. The linear model is appropriate since most residual assumptions are met, and the extreme point represents an explainable special cause. (correct answer)
  2. The linear model should be rejected because any residual exceeding 3 standard deviations indicates fundamental model inadequacy.
  3. The linear model is inappropriate because the extreme residual violates the normality assumption required for valid statistical inference.
  4. The linear model requires modification by adding indicator variables to account for maintenance periods and other operational factors.
Explanation: Since the extreme residual has an explainable special cause (machine servicing) and other assumptions are met, the linear model is appropriate for normal operating conditions. The outlier represents an unusual circumstance rather than model inadequacy. Choice B is too rigid about outlier thresholds. Choice C misunderstands that one outlier doesn't necessarily violate normality. Choice D overcomplicates by suggesting model modifications when the special cause is identifiable and removable.

Question 4

A sports analyst models the relationship between player height and scoring average using linear regression. The analysis shows R2=0.48R^2 = 0.48, randomly scattered residuals, and normal distribution of errors. A colleague argues the model is inadequate because R2<0.50R^2 < 0.50 means the model explains less than half the variance. How should the analyst respond to this criticism?

  1. The colleague is correct; models with R2<0.50R^2 < 0.50 are inadequate for any practical purpose and should be abandoned completely.
  2. The colleague is incorrect; R2=0.48R^2 = 0.48 indicates a moderately strong relationship, and model adequacy depends on context and residual behavior. (correct answer)
  3. The colleague is partially correct; while R2=0.48R^2 = 0.48 is low, the model can still be useful if additional predictors are added.
  4. The colleague is incorrect; R2R^2 values are irrelevant for model evaluation when residuals show appropriate patterns and normal distribution.
Explanation: Model adequacy cannot be judged solely by R² thresholds. In sports analytics, R² = 0.48 represents a moderately strong relationship, and the good residual properties suggest the linear model is appropriate for the data. Context matters—human performance has many sources of variability. Choice A applies an arbitrary and too-strict threshold. Choice C unnecessarily accepts the premise that 0.48 is inadequate. Choice D goes too far by dismissing R² entirely.

Question 5

A pharmaceutical company models drug concentration in blood versus time using linear regression on the logarithm of concentration. The residual plot shows random scatter around zero with constant variance. However, when examining the original (non-transformed) data, the relationship appears curved. What should be concluded about model appropriateness?

  1. The linear model is inappropriate because the curved pattern in original data indicates fundamental nonlinearity that cannot be addressed by transformation.
  2. The linear model is appropriate because good residual behavior after log transformation indicates the model successfully captures the underlying relationship. (correct answer)
  3. The linear model is inappropriate because transformations artificially force linearity and mask the true nonlinear nature of the drug concentration process.
  4. The linear model requires validation by comparing both transformed and untransformed residual patterns before determining final model appropriateness.
Explanation: Log transformation is appropriate for exponential decay processes like drug concentration. The good residual behavior after transformation indicates the linear model on log scale successfully captures the relationship. The curvature in original data is expected and properly addressed by transformation. Choice A misunderstands appropriate use of transformations. Choice C incorrectly suggests transformations are artificial rather than theoretically justified. Choice D is unnecessarily complex when the transformed residuals clearly show good behavior.

Question 6

A market researcher models the relationship between advertising spend and sales revenue using linear regression. The analysis yields R2=0.83R^2 = 0.83 and the residual plot shows random scatter. However, the researcher notices that the model's predictions become less accurate for advertising spends above $50,000, even though these represent only 8% of the data. What is the most appropriate evaluation of model adequacy?

  1. The model is adequate because R2=0.83R^2 = 0.83 and random residuals indicate good overall fit, and 8% of problematic data is negligible.
  2. The model is inadequate because any systematic prediction errors, regardless of frequency, indicate fundamental problems with model specification.
  3. The model is inadequate because poor performance in the high-spend range suggests nonlinear effects that invalidate the entire linear approach.
  4. The model is adequate for most situations but limitations should be noted for high advertising spend predictions, possibly requiring separate analysis. (correct answer)
Explanation: When evaluating regression models, you need to balance overall performance metrics with practical limitations and specific use cases. A high R2R^2 value and random residuals indicate good general model performance, but they don't tell the complete story about model adequacy across all data ranges. The correct approach is answer D because it recognizes that model evaluation should be nuanced and context-dependent. With R2=0.83R^2 = 0.83, the model explains 83% of variance and performs well for the majority of data. However, acknowledging poor performance in the high-spend range (above $50,000) is crucial for responsible model use. This approach allows you to use the model where it's reliable while flagging its limitations for specific scenarios. Answer A is wrong because dismissing systematic errors in any data range, even if small, ignores important practical implications. High advertising spends likely represent significant business decisions where prediction accuracy is critical. Answer B is too extreme—it assumes any prediction errors indicate complete model failure. In practice, models often have limitations in specific ranges while remaining useful overall. Answer C incorrectly concludes that localized nonlinear effects "invalidate the entire linear approach." Poor performance in one range doesn't necessarily mean the linear model is inappropriate for the majority of data where it performs well. Study tip: Remember that model adequacy isn't binary—models can be useful with limitations. Always consider both statistical metrics and practical context, especially noting where models might fail in business-critical scenarios.

Question 7

A data analyst fits a linear model to predict house prices based on square footage. The residual plot shows that 95% of residuals fall within 2 standard deviations of zero, and residuals appear randomly distributed. However, when examining the original scatterplot more carefully, three houses with square footage above 4000 sq ft have much higher prices than the linear trend would predict. What is the most appropriate conclusion?

  1. The linear model is appropriate because the residual plot indicates random errors and normal distribution of residuals around the regression line.
  2. The linear model is inappropriate because the three high-value outliers suggest a nonlinear relationship that requires polynomial regression instead.
  3. The linear model may be appropriate for most data, but influential outliers should be investigated as they could affect model reliability and predictions. (correct answer)
  4. The linear model is inappropriate because any data points that deviate significantly from the trend line indicate fundamental model inadequacy.
Explanation: While the residual plot suggests overall good fit, the presence of influential outliers (high-value houses) requires careful consideration as they could disproportionately affect the model. The analyst should investigate these points rather than immediately accepting or rejecting the model. Choice A ignores the outlier issue. Choice B jumps to conclusions about nonlinearity without sufficient evidence. Choice D is too extreme—outliers don't automatically invalidate a model but need investigation.

Question 8

A researcher examining plant growth versus sunlight hours obtains these results: correlation r=0.95r = 0.95, R2=0.90R^2 = 0.90, and the residual plot appears random with constant variance. However, the researcher realizes the data includes measurements from both indoor and outdoor plants, with outdoor plants showing consistently higher growth rates at all sunlight levels. How does this affect model appropriateness?

  1. The model remains appropriate since excellent correlation and R2R^2 values demonstrate strong predictive power despite the plant location differences.
  2. The model becomes inappropriate because mixing different plant populations violates the independence assumption required for valid linear regression.
  3. The model is inappropriate because the systematic difference between indoor and outdoor plants indicates heteroscedasticity in the error structure.
  4. The model is appropriate for overall trends but may mask important differences between plant types that could improve prediction accuracy. (correct answer)
Explanation: When evaluating regression models, you need to consider both statistical measures and potential lurking variables that could affect your conclusions. Strong correlation and R2R^2 values indicate the model fits the data well, but they don't tell the whole story about whether you're missing important patterns. The correct answer is D because while the model shows strong overall predictive power, it's treating indoor and outdoor plants as one homogeneous group when they clearly behave differently. The systematic difference between plant types suggests you could build a better model by accounting for location (perhaps through separate regression lines or including location as a categorical variable). The current model works for general predictions but overlooks meaningful biological differences. Option A is wrong because excellent statistical measures don't automatically mean the model is optimal—you can have good fit while still missing important subgroup patterns. Option B incorrectly identifies this as an independence violation; the plants' measurements are still independent observations, just from different populations. The independence assumption refers to errors being uncorrelated, not to having homogeneous subjects. Option C misidentifies the issue as heteroscedasticity—the residual plot actually shows constant variance, so the error variance is consistent. When you see regression problems with high correlation but mention of different groups or conditions, ask yourself: "Could I improve predictions by treating these groups separately?" Strong overall statistics can mask important subgroup differences that, when properly modeled, lead to better insights and predictions.

Question 9

A researcher collected data on the relationship between hours of study per week and test scores. After fitting a linear regression model, the residual plot shows residuals that are positive for low and high fitted values but negative for middle fitted values, forming a clear U-shaped pattern. Additionally, the scatterplot shows R2=0.75R^2 = 0.75. What can be concluded about the appropriateness of the linear model?

  1. The linear model is appropriate because the high R2R^2 value indicates strong correlation between variables and good model fit.
  2. The linear model is inappropriate because the U-shaped residual pattern indicates a systematic nonlinear relationship that violates linearity assumptions. (correct answer)
  3. The linear model is appropriate because residuals showing both positive and negative values demonstrate that errors are randomly distributed around zero.
  4. The linear model is inappropriate because the R2R^2 value of 0.75 is too low to justify using linear regression for prediction purposes.
Explanation: The U-shaped pattern in residuals indicates systematic deviation from linearity, suggesting the relationship is curved rather than linear. Even with a high R² value, this pattern violation makes the linear model inappropriate. Choice A incorrectly focuses only on R² while ignoring the residual pattern. Choice C misinterprets what constitutes random residuals. Choice D incorrectly suggests R² = 0.75 is too low when it's actually quite strong.

Question 10

Two students analyze the same dataset using linear regression. Student A reports that the linear model is appropriate based on R2=0.81R^2 = 0.81 and a correlation of r=0.90r = 0.90. Student B argues the model is inappropriate after examining residuals and finding that 80% of negative residuals occur in the first half of the ordered data, while 80% of positive residuals occur in the second half. Who is correct and why?

  1. Student B is correct because the residual pattern indicates systematic trend that violates linear model assumptions (correct answer)
  2. Student A is correct because high R2R^2 and correlation definitively prove linear relationship appropriateness
  3. Both students are partially correct, but Student A's evidence is more reliable for model evaluation
  4. Neither student has sufficient evidence because model appropriateness requires examining prediction accuracy on new data
Explanation: When evaluating linear regression models, you need to examine both goodness-of-fit statistics AND residual patterns. High R2R^2 and correlation values only tell you how much variation the model explains, not whether the linear assumption is valid. Student B has identified a classic violation of linear regression assumptions. When residuals show systematic patterns—like negative residuals clustering in the first half of ordered data and positive residuals clustering in the second half—this indicates the relationship isn't truly linear. This pattern suggests the data follows a curved relationship that the straight line can't properly capture, even though it might still explain a high percentage of variance. Choice A is correct because residual analysis is the proper diagnostic tool for checking model assumptions. The systematic residual pattern Student B found indicates the linear model is fundamentally inappropriate, regardless of the high R2R^2. Choice B is wrong because R2=0.81R^2 = 0.81 and r=0.90r = 0.90 don't prove linear appropriateness—they only show the line fits well overall. A curved relationship can still produce high correlation with a linear model while violating assumptions. Choice C incorrectly prioritizes fit statistics over assumption checking. Student A's evidence is actually less reliable because it ignores whether the linear model is appropriate in the first place. Choice D misses the point entirely. While prediction accuracy matters, you must first ensure your model meets basic assumptions before trusting any predictions. Remember: Always examine residual plots before trusting any regression statistics. Patterns in residuals reveal assumption violations that summary statistics can mask.

Question 11

An analyst fits a linear model to predict monthly sales based on advertising spend. The residual analysis reveals that residuals are approximately normally distributed around zero with constant variance, but there is a clear time pattern: residuals alternate between positive and negative values in consecutive months. What does this suggest about the linear model?

  1. The linear model misses an important time-related component that affects sales beyond advertising spend (correct answer)
  2. The linear model is appropriate since residuals meet the normality and constant variance assumptions
  3. The linear model is inappropriate because alternating residuals indicate measurement error in the data
  4. The linear model is appropriate but would benefit from removing the months with negative residuals
Explanation: When evaluating linear regression models, you need to check multiple assumptions: normality of residuals, constant variance (homoscedasticity), and independence of residuals. While the first two assumptions appear satisfied here, the clear time pattern in residuals signals a violation of the independence assumption. The alternating positive-negative pattern in consecutive months indicates serial correlation or autocorrelation in the residuals. This means the model is systematically missing something that varies predictably over time. Perhaps sales have seasonal trends, momentum effects from previous months, or other temporal dynamics that advertising spend alone cannot capture. The model is underfitting by ignoring this time-related component. Looking at the wrong answers: Option B incorrectly focuses only on normality and variance while ignoring the crucial independence assumption. The time pattern is a red flag that cannot be dismissed. Option C misidentifies the issue as measurement error when it's actually a specification problem—the model structure itself is incomplete. Option D suggests an inappropriate data manipulation approach that would worsen the problem by creating gaps in the time series and wouldn't address the underlying issue. The alternating residual pattern is the model's way of telling you it's missing systematic time-related information that could improve predictions. Study tip: When analyzing residuals, always plot them against time (if applicable) and look for patterns. Random residuals should show no discernible trends or cycles—any systematic pattern suggests your model is incomplete and needs additional variables or terms.

Question 12

A researcher collected data on the relationship between study hours per week and test scores for 25 students. After fitting a linear regression model, the residual plot shows a clear U-shaped pattern with most residuals negative in the middle range of study hours and positive at both extremes. Based on the residual plot described above, which conclusion about the linear model is most appropriate?

  1. The linear model is appropriate because the residuals are randomly distributed around zero
  2. The linear model is inappropriate because the residual pattern suggests a nonlinear relationship exists (correct answer)
  3. The linear model is appropriate but has some outliers that should be removed from the dataset
  4. The linear model is inappropriate because the residuals show increasing variance across the range
Explanation: A U-shaped pattern in residuals indicates systematic deviation from linearity, suggesting the true relationship is curved (possibly quadratic). This violates the assumption of linearity. Choice A is wrong because random distribution would show no pattern. Choice C is wrong because this describes a systematic pattern, not isolated outliers. Choice D is wrong because this describes heteroscedasticity (fan pattern), not the U-shaped pattern described.

Question 13

A student fits a linear model to data and obtains R2=0.64R^2 = 0.64. When examining the residual plot, they notice that residuals tend to be positive for low fitted values, negative for middle fitted values, and positive again for high fitted values. The student concludes that since R2R^2 is reasonably high, the linear model is adequate. What is wrong with this reasoning?

  1. The R2R^2 value is too low to indicate a good fit, regardless of the residual pattern
  2. R2R^2 is not relevant for model evaluation; only the residual pattern matters for determining appropriateness
  3. The reasoning is correct because R2>0.5R^2 > 0.5 always indicates an appropriate linear model
  4. R2R^2 measures association strength but not whether the relationship is actually linear; the residual pattern suggests nonlinearity (correct answer)
Explanation: When evaluating regression models, you need to assess both the strength of the relationship and whether your model form is appropriate. This requires looking at multiple diagnostic tools together, not relying on any single measure. The correct answer is D because R2R^2 only tells you how much variance your model explains, not whether a linear relationship is the right choice. An R2=0.64R^2 = 0.64 means 64% of the variation is captured, but the residual pattern described—positive, then negative, then positive again across fitted values—forms a clear curved pattern. This indicates the true relationship is nonlinear, and forcing a linear model creates systematic errors. You could have a high R2R^2 with a completely inappropriate model if the underlying relationship is strong but curved. Choice A is wrong because R2=0.64R^2 = 0.64 isn't inherently "too low"—the threshold for "good" depends on context. The real problem isn't the magnitude but what's missing from this analysis. Choice B incorrectly dismisses R2R^2 entirely; both metrics matter for different reasons. Choice C represents a dangerous misconception—no single R2R^2 value automatically indicates model appropriateness regardless of other diagnostics. The key insight is that this curved residual pattern is a classic sign you need a nonlinear model (perhaps quadratic). A proper linear model should show randomly scattered residuals with no systematic patterns. Study tip: Always examine residual plots alongside R2R^2. High R2R^2 with patterned residuals typically means you need a different model form, not that your current model is adequate.

Question 14

A linear regression was performed to predict house prices based on square footage. The correlation coefficient is r=0.85r = 0.85, and the residual plot shows points scattered randomly around zero with roughly constant spread. However, one house with 3000 sq ft sold for $50,000 less than predicted. What is the most reasonable interpretation of these results?

  1. The linear model is inappropriate because the correlation is too weak to be meaningful
  2. The linear model is appropriate, and the unusual house price confirms that square footage perfectly predicts house values
  3. The linear model is inappropriate because the large residual indicates systematic bias in the model
  4. The linear model is appropriate overall, but this particular house may have unique characteristics not captured by square footage alone (correct answer)
Explanation: When evaluating linear regression models, you need to consider multiple pieces of evidence together: the correlation coefficient, residual patterns, and individual outliers. A strong model can still have occasional unusual data points. The correlation coefficient of r=0.85r = 0.85 indicates a strong positive linear relationship between square footage and house price. The randomly scattered residuals with constant spread confirm that the linear model assumptions are met - there's no systematic pattern suggesting the model is fundamentally flawed. However, one house deviating significantly from predictions doesn't invalidate the entire model. Option D correctly recognizes that the linear model is working well overall, but acknowledges that house prices depend on factors beyond just square footage. This particular house likely has unique characteristics (poor condition, undesirable location, foreclosure sale, etc.) that the simple model cannot account for. Option A incorrectly suggests that r=0.85r = 0.85 is weak, when it actually represents a strong correlation. Option B makes the impossible claim that square footage "perfectly" predicts house values - no real-world relationship is perfect, and one outlier doesn't confirm perfect prediction anyway. Option C misinterprets a single outlier as evidence of systematic bias, but systematic bias would appear as patterns in the residual plot, not random scatter. Remember that in regression analysis, individual outliers don't necessarily invalidate a model. Look at the overall pattern: strong correlation plus random residuals typically indicates a useful model, even when some data points don't fit perfectly.

Question 15

A researcher studying the relationship between temperature and ice cream sales creates both a scatterplot and residual plot. The scatterplot shows a strong positive linear relationship (r=0.92r = 0.92), but the residual plot reveals that all residuals for temperatures below 70°F are positive, while all residuals for temperatures above 80°F are negative. What does this pattern most likely indicate?

  1. The linear model is excellent because the high correlation coefficient indicates strong predictive power
  2. The data contains measurement errors that should be corrected before fitting any model
  3. The linear model systematically underestimates sales at low temperatures and overestimates at high temperatures (correct answer)
  4. The linear model is inappropriate due to the presence of outliers in the temperature data
Explanation: When residuals are consistently positive in one range and negative in another, this indicates systematic bias - the model consistently predicts too low in one region and too high in another. This suggests the true relationship may have a different slope in different regions or be curved. Choice A ignores the systematic residual pattern. Choice B assumes measurement error without evidence. Choice D misidentifies systematic bias as outliers.