Biostatistics Quiz: Slope Intercept R2 And Residuals
20 questions · exam conditions
0:00
Slope Intercept R2 And ResidualsQuestion 1 of 20

In a study of bone density, researchers obtain the regression equation y^=1.8−0.03x\hat{y} = 1.8 - 0.03x where yy is bone density (g/cm²) and xx is age (years). For a 50-year-old participant, the predicted bone density is 0.30 g/cm² and the observed value is 0.42 g/cm². What is the residual for this participant, and what does it indicate?

+0.12+0.12 g/cm², indicating the participant's bone density is higher than predicted by the model
−0.12-0.12 g/cm², indicating the model overestimated this participant's bone density relative to their age
+0.12+0.12 g/cm², indicating a positive correlation between age and bone density for this individual
−0.30-0.30 g/cm², indicating the participant's bone density deviates significantly from the population mean
← Back to quizzes

Biostatistics Quiz

Biostatistics Quiz: Slope Intercept R2 And Residuals

Practice Slope Intercept R2 And Residuals in Biostatistics 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 Slope Intercept R2 And Residuals, giving you a quick way to practice the rules, question types, and explanations that matter most for Biostatistics.

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

In a study of bone density, researchers obtain the regression equation y^=1.8−0.03x\hat{y} = 1.8 - 0.03x where yy is bone density (g/cm²) and xx is age (years). For a 50-year-old participant, the predicted bone density is 0.30 g/cm² and the observed value is 0.42 g/cm². What is the residual for this participant, and what does it indicate?

  1. +0.12+0.12 g/cm², indicating the participant's bone density is higher than predicted by the model (correct answer)
  2. −0.12-0.12 g/cm², indicating the model overestimated this participant's bone density relative to their age
  3. +0.12+0.12 g/cm², indicating a positive correlation between age and bone density for this individual
  4. −0.30-0.30 g/cm², indicating the participant's bone density deviates significantly from the population mean
Explanation: Residual = observed - predicted = 0.42 - 0.30 = +0.12 g/cm². A positive residual means the observed value is above the predicted value. Choice B has the wrong sign. Choice C misinterprets what a residual indicates (it's not about correlation). Choice D incorrectly calculates the residual and misinterprets its meaning.

Question 2

A researcher studies the relationship between hours of sleep per night (X) and test scores (Y) in college students. The regression equation is Y^=45+8X\hat{Y} = 45 + 8X. If a student who normally sleeps 6 hours per night increases their sleep to 8 hours per night, what is the predicted change in their test score?

  1. The test score is predicted to increase by 8 points
  2. The test score is predicted to increase by 16 points (correct answer)
  3. The test score is predicted to increase by 109 points total
  4. The test score is predicted to increase by 2 points
  5. The test score is predicted to increase by 45 points
Explanation: When you encounter a linear regression problem asking about predicted changes, focus on the slope coefficient—it tells you how much Y changes for each one-unit increase in X. In the regression equation Y^=45+8X\hat{Y} = 45 + 8X, the slope is 8, meaning each additional hour of sleep predicts an 8-point increase in test score. The student increases sleep from 6 to 8 hours, which is a 2-hour increase. Since each hour predicts an 8-point improvement, the total predicted change is 2×8=162 \times 8 = 16 points. You can verify this by calculating both predicted scores: at 6 hours, Y^=45+8(6)=93\hat{Y} = 45 + 8(6) = 93; at 8 hours, Y^=45+8(8)=109\hat{Y} = 45 + 8(8) = 109. The difference is 109−93=16109 - 93 = 16 points, confirming answer B. Choice A incorrectly uses just the slope coefficient (8) without accounting for the 2-hour change. This would only be correct if sleep increased by exactly 1 hour. Choice C gives the total predicted score at 8 hours of sleep (109 points) rather than the change in score—a common trap when students calculate the final value instead of the difference. Choice D simply uses the change in hours of sleep (2) without applying the slope coefficient. Remember: in regression problems asking for predicted change, multiply the change in X by the slope coefficient. Don't confuse the change in the outcome with the total predicted value, and always account for the full magnitude of change in the predictor variable.

Question 3

In a regression analysis of blood pressure (Y) versus age (X), the equation is Y^=95+1.2X\hat{Y} = 95 + 1.2X with R2=0.64R^2 = 0.64. What does the intercept value of 95 represent in this context?

  1. The predicted blood pressure for a person aged 95 years
  2. The predicted blood pressure for a hypothetical person aged 0 years (correct answer)
  3. The average blood pressure across all ages in the study
  4. The maximum possible blood pressure predicted by the model
  5. The blood pressure increase for each additional year of age
Explanation: When interpreting linear regression equations, you need to understand what each component represents. In the equation Y^=95+1.2X\hat{Y} = 95 + 1.2X, this follows the standard form Y^=a+bX\hat{Y} = a + bX, where a is the y-intercept and b is the slope. The intercept represents the predicted value of Y when X equals zero. In this blood pressure study, when age (X) = 0, the predicted blood pressure is 95 mmHg. This is the theoretical blood pressure for a newborn, making option B correct. Looking at the incorrect options: Option A misinterprets 95 as an age value rather than a blood pressure prediction. The intercept has nothing to do with a 95-year-old person. Option C confuses the intercept with the mean of Y values. While related, the intercept isn't necessarily the average blood pressure across all ages in your dataset—it's specifically the predicted value at age zero. Option D incorrectly suggests the intercept represents a maximum value. The intercept is just one point on the regression line; depending on the data range and slope, other predicted values could be higher or lower. Note that while the intercept has a clear mathematical interpretation, it may not always be clinically meaningful (newborns don't typically have blood pressure of 95 mmHg). However, the statistical definition remains the same regardless of practical relevance. Study tip: Remember the intercept formula: set X = 0 in your regression equation. The resulting Y value is always your intercept's meaning, regardless of whether X = 0 makes real-world sense.

Question 4

A study examines the relationship between exercise hours per week (X) and weight loss in pounds (Y). The regression gives R2=0.36R^2 = 0.36. A colleague claims this means exercise causes 36% of all weight loss. What is wrong with this interpretation?

  1. R2R^2 measures association strength, not causation, and represents proportion of variance explained (correct answer)
  2. The value should be squared to get the true percentage of causation
  3. R2=0.36R^2 = 0.36 means only 36% of people in the study lost weight from exercise
  4. The interpretation is correct but applies only to this specific sample
  5. R2R^2 values below 0.5 indicate no meaningful relationship exists between variables
Explanation: When you encounter questions about R2R^2 in regression analysis, remember that this statistic measures the strength of association between variables, not causation. Understanding this distinction is crucial for interpreting statistical results correctly. The correct interpretation of R2=0.36R^2 = 0.36 is that 36% of the variance in weight loss (Y) is explained by the variance in exercise hours (X). This tells you how well your regression line fits the data, but says nothing about whether exercise actually causes weight loss. Correlation does not imply causation - there could be confounding variables, reverse causation, or the relationship might be purely coincidental. Answer A correctly identifies that R2R^2 measures association strength and represents the proportion of variance explained, not causation. Answer B is incorrect because R2R^2 is already interpreted as a proportion - you don't square it again. The value 0.36 directly means 36% of variance is explained. Answer C misinterprets R2R^2 entirely - it doesn't tell you what percentage of people lost weight, but rather how much of the total variation in weight loss amounts is accounted for by exercise hours. Answer D incorrectly accepts the causation claim, suggesting the interpretation is valid for this sample, which perpetuates the fundamental error of confusing association with causation. Study tip: Always remember that R2R^2 = "percentage of variance explained by the model," never "percentage of causation." Statistical association alone cannot establish causal relationships - that requires experimental design or additional analytical methods.

Question 5

In a regression of medication dosage (X) versus symptom relief score (Y), one patient received 10mg and had a symptom relief score of 75. The predicted score for this dosage was 70. What is the residual for this patient?

  1. The residual is -5, indicating the model overestimated this patient's relief
  2. The residual is 5, indicating the model underestimated this patient's relief (correct answer)
  3. The residual is 75, which is the patient's actual symptom relief score
  4. The residual is 70, which is the predicted symptom relief score
  5. The residual is 10, which corresponds to the medication dosage given
Explanation: When you encounter regression residual questions, you're dealing with the fundamental concept of how well a model fits individual data points. A residual measures the difference between what actually happened and what the model predicted would happen. The residual is calculated as: Residual=Observed value−Predicted value\text{Residual} = \text{Observed value} - \text{Predicted value}. In this case, the patient's actual symptom relief score was 75, and the model predicted 70, so the residual is 75−70=575 - 70 = 5. Since this residual is positive, it means the actual outcome was higher than predicted—the model underestimated this patient's symptom relief. This confirms answer B is correct. Answer A makes the calculation error of subtracting in the wrong direction (predicted minus observed instead of observed minus predicted), giving -5, and misinterprets what this would mean. Answer C confuses the residual with the raw observed value—the residual isn't the patient's actual score, but rather how far off the prediction was. Answer D similarly confuses the residual with the predicted value itself, missing the entire concept that residuals measure prediction error. Remember this pattern: positive residuals mean the model underestimated (actual > predicted), while negative residuals mean the model overestimated (actual < predicted). The key formula is always observed minus predicted, and the sign tells you the direction of the error. This concept appears frequently in biostatistics when evaluating how well models perform.

Question 6

Two researchers study the same dataset relating study time (X) to exam scores (Y). Researcher A reports R2=0.49R^2 = 0.49 while Researcher B reports R2=0.25R^2 = 0.25. If both used valid regression techniques, what could explain this difference?

  1. Researcher A made a calculation error since R2R^2 values should be identical
  2. Researcher A included additional predictor variables in a multiple regression model (correct answer)
  3. Researcher B used a different correlation coefficient calculation method than Researcher A
  4. The difference represents sampling error and both values are estimates of the true R2R^2
  5. Researcher A analyzed the data in reverse with exam scores predicting study time
Explanation: When you encounter questions about different R2R^2 values from the same dataset, think about what factors can legitimately change the proportion of variance explained in regression analysis. Researcher B's R2=0.25R^2 = 0.25 likely comes from simple linear regression using only study time (X) to predict exam scores (Y). This means study time explains 25% of the variance in exam scores. Researcher A's higher R2=0.49R^2 = 0.49 suggests they used multiple regression, adding other predictor variables beyond just study time. When you include additional relevant predictors (like prior GPA, sleep hours, or attendance), R2R^2 typically increases because you're explaining more of the outcome's variance. This is why answer B is correct. Answer A is wrong because R2R^2 values don't have to be identical when using different regression approaches on the same data. Answer C misunderstands the situation—both researchers would calculate the same correlation coefficient between study time and exam scores, but R2R^2 measures explained variance in regression models, not just correlation. Answer D incorrectly suggests sampling error, but both researchers used the same complete dataset, so there's no sampling involved. The key insight is that R2R^2 depends on your model specification. Simple regression (one predictor) will generally have lower R2R^2 than multiple regression (several predictors) because additional variables usually explain additional variance. Remember: R2R^2 measures how much variance your chosen predictors explain, so different predictor combinations yield different R2R^2 values.

Question 7

In a regression of height (X) versus weight (Y), the slope is 2.3 pounds per inch. A person who is 68 inches tall weighs 160 pounds, while the regression predicts 155 pounds. If this person grows to 70 inches and their weight increases proportionally to maintain the same residual, what will their new weight be?

  1. 164.6 pounds, maintaining the same 5-pound positive residual above prediction (correct answer)
  2. 160 pounds, keeping their weight constant despite height change
  3. 159.6 pounds, calculated as the new predicted value only
  4. 164.0 pounds, adding the height change times slope to current weight
  5. 170 pounds, using the new height value as the weight estimate
Explanation: When you encounter regression problems involving residuals and proportional changes, you need to understand three key components: the regression line (predicted values), actual values, and residuals (the difference between them). Let's work through this systematically. First, calculate the current residual: the person weighs 160 pounds but the regression predicts 155 pounds, so their residual is 160−155=+5160 - 155 = +5 pounds. This means they weigh 5 pounds more than the regression line predicts. When the person grows to 70 inches, the new predicted weight becomes: 155+(2 inches×2.3 pounds/inch)=155+4.6=159.6155 + (2 \text{ inches} \times 2.3 \text{ pounds/inch}) = 155 + 4.6 = 159.6 pounds. Since their weight increases "proportionally to maintain the same residual," they'll still weigh 5 pounds above prediction: 159.6+5=164.6159.6 + 5 = 164.6 pounds. Answer A correctly identifies this as 164.6 pounds with the same 5-pound positive residual. Answer B incorrectly assumes weight stays constant at 160 pounds, which would actually change the residual from +5 to +0.4 pounds. Answer C gives only the predicted value (159.6 pounds) without adding back the residual, resulting in a -5 pound residual instead. Answer D incorrectly adds the slope change to the current actual weight (160 + 4.6 = 164.6) but arrives at the right number through wrong reasoning. Remember: "maintaining the same residual" means the difference between actual and predicted values stays constant, not the actual weight itself. Always calculate the new prediction first, then apply the original residual.

Question 8

A medical study reports that medication dosage (X) explains 64% of the variance in recovery time (Y), with the equation Y^=15−0.8X\hat{Y} = 15 - 0.8X. If a patient receives 10mg of medication, what is the predicted recovery time?

  1. 7 days, calculated by substituting X = 10 into the regression equation (correct answer)
  2. 15 days, which represents the baseline recovery time without medication
  3. 8 days, calculated by subtracting the slope from the dosage amount
  4. 6.4 days, calculated by multiplying the dosage by the explained variance
  5. 23 days, calculated by adding the slope times dosage to the intercept
Explanation: When you encounter a linear regression equation, you're looking at a predictive model where you can substitute any X value to find the corresponding predicted Y value. The equation Y^=15−0.8X\hat{Y} = 15 - 0.8X tells you exactly how to calculate the predicted recovery time for any given medication dosage. To find the predicted recovery time for a 10mg dosage, you simply substitute X = 10 into the equation: Y^=15−0.8(10)=15−8=7\hat{Y} = 15 - 0.8(10) = 15 - 8 = 7 days. This straightforward substitution gives you the answer. Looking at the wrong choices: Option B (15 days) represents the y-intercept, which would be the predicted recovery time when X = 0 (no medication), not when X = 10. Option C (8 days) makes the error of simply subtracting the slope (-0.8) from the dosage (10), ignoring the regression equation entirely. Option D (6.4 days) incorrectly multiplies the dosage by the R² value (0.64), confusing the coefficient of determination with the actual prediction calculation. The 64% variance explanation and R² value are red herrings in this problem—they tell you how well the model fits the data but aren't needed for making predictions. Study tip: In regression problems, always use the given equation for predictions by direct substitution. Don't let additional statistical information like R² distract you from the basic calculation. The regression equation is your prediction tool—everything else is just context about model quality.

Question 9

In a regression of temperature (°F) versus ice cream sales ($), the intercept is -200 dollars. What does this negative intercept most likely represent?

  1. The model predicts negative sales at 0°F, indicating extrapolation beyond reasonable data range (correct answer)
  2. There is a calculation error since intercepts cannot be negative in regression analysis
  3. Ice cream sales decrease by $200 for each degree increase in temperature
  4. The correlation coefficient between temperature and sales must also be negative
  5. The regression line has a negative slope, indicating decreased sales with higher temperature
Explanation: When interpreting regression intercepts, you need to understand what the intercept represents: the predicted y-value when x equals zero. In this temperature-sales regression, the intercept tells you the predicted ice cream sales when the temperature is 0°F. A negative intercept of -$200 means the model predicts you'd have negative $200 in sales at 0°F. While mathematically valid, this is practically meaningless—you can't have negative sales. This occurs because 0°F likely falls far outside the range of temperatures in your original data. Most ice cream sales data probably comes from warmer months (say, 60-90°F), so predicting what happens at 0°F requires dangerous extrapolation beyond your data's scope. Looking at the wrong answers: Option B is incorrect because intercepts can absolutely be negative—there's no mathematical restriction preventing this. Option C confuses the intercept with the slope; the slope (not intercept) tells you how much y changes per unit change in x. Option D makes a false connection—while this regression likely has a positive correlation (more heat = more ice cream), the sign of the intercept doesn't determine the correlation's sign. Study tip: When you see unusual intercept values (especially negative ones in contexts where negative y-values don't make sense), immediately think "extrapolation problem." The intercept is only meaningful if x = 0 falls within your data's reasonable range. Always consider whether the x = 0 point makes practical sense in your study's context.

Question 10

In a study of plant growth, the regression of height (Y) on sunlight hours (X) yields Y^=2.8+1.5X\hat{Y} = 2.8 + 1.5X with R2=0.49R^2 = 0.49. A plant receiving 6 hours of sunlight daily has an actual height of 13.2 cm. How does this plant's growth compare to the model's prediction?

  1. The plant grew exactly as predicted since the residual equals zero
  2. The plant grew 1.4 cm more than predicted, showing above-average response to sunlight (correct answer)
  3. The plant grew 1.4 cm less than predicted, showing below-average response to sunlight
  4. The plant's growth cannot be compared without knowing the correlation coefficient
  5. The plant grew 13.2 cm total, which represents excellent agreement with the model
Explanation: When you encounter regression problems asking you to compare actual versus predicted values, you're working with residuals - the difference between what actually happened and what the model predicted would happen. First, calculate the predicted height using the regression equation. With Y^=2.8+1.5X\hat{Y} = 2.8 + 1.5X and X=6X = 6 hours of sunlight: Y^=2.8+1.5(6)=2.8+9=11.8\hat{Y} = 2.8 + 1.5(6) = 2.8 + 9 = 11.8 cm. The model predicts this plant should be 11.8 cm tall. The residual equals actual minus predicted: 13.2−11.8=1.413.2 - 11.8 = 1.4 cm. Since this is positive, the plant grew 1.4 cm more than the model predicted, indicating above-average response to sunlight. Answer A is wrong because the residual clearly doesn't equal zero - there's a 1.4 cm difference between actual and predicted height. Answer C incorrectly states the plant grew less than predicted, but since the actual height (13.2 cm) exceeds the predicted height (11.8 cm), the plant actually performed better than expected. Answer D is incorrect because you don't need the correlation coefficient to calculate residuals - you only need the regression equation and the actual observed value. Remember that positive residuals mean "better than predicted" while negative residuals mean "worse than predicted." Always calculate predicted value first, then subtract it from the actual value to find the residual and determine whether performance exceeded or fell short of expectations.

Question 11

A regression equation Y^=120−3.5X\hat{Y} = 120 - 3.5X predicts blood pressure (Y) from hours of weekly exercise (X). What is the most appropriate interpretation of the negative slope?

  1. The model is invalid because biological variables cannot have negative relationships
  2. For each additional hour of weekly exercise, blood pressure decreases by 3.5 units on average (correct answer)
  3. People who exercise exactly 3.5 hours per week will have zero blood pressure
  4. The intercept should be negative to match the negative slope direction
  5. Blood pressure reaches its minimum value when exercise equals 120 hours per week
Explanation: When interpreting linear regression equations, focus on what the slope coefficient tells you about the relationship between your predictor and outcome variables. The regression equation Y^=120−3.5X\hat{Y} = 120 - 3.5X follows the standard form Y^=b0+b1X\hat{Y} = b_0 + b_1X, where the slope b1=−3.5b_1 = -3.5 quantifies how much Y changes for each one-unit increase in X. The slope of -3.5 means that for every additional hour of weekly exercise, the predicted blood pressure decreases by 3.5 units on average. This makes biological sense—exercise typically lowers blood pressure. Answer B correctly captures this interpretation. Answer A reflects a fundamental misunderstanding. Negative relationships are common and valid in biology and medicine. Many variables have inverse relationships, like exercise and blood pressure, or smoking cessation and lung cancer risk. Answer C confuses the slope with a specific prediction point and misinterprets what happens at X = 3.5. When X = 3.5, Y^=120−3.5(3.5)=107.75\hat{Y} = 120 - 3.5(3.5) = 107.75, not zero. Answer D incorrectly suggests the intercept and slope must have matching signs. The intercept (120) represents the predicted blood pressure when exercise hours equal zero, which is independent of the slope's direction. Remember that in regression, the slope always tells you the average change in the outcome variable per one-unit increase in the predictor variable. The sign indicates direction: positive slopes show increases together, negative slopes show one increases as the other decreases.

Question 12

A regression analysis shows R2=0.81R^2 = 0.81 for predicting cholesterol level from dietary fat intake. If the correlation coefficient r is negative, what is the value of r?

  1. r=−0.81r = -0.81 because R2R^2 and r have the same sign
  2. r=−0.9r = -0.9 because r is the negative square root of R2R^2 (correct answer)
  3. r=0.9r = 0.9 because R2R^2 is always the square of the absolute value
  4. r=−0.656r = -0.656 because this equals −R2-R^2 divided by the square root
  5. r=0.81r = 0.81 because R2R^2 represents the same relationship strength
Explanation: When you encounter questions about the relationship between the correlation coefficient (r) and the coefficient of determination (R2R^2), remember that R2R^2 is always the square of r, regardless of whether r is positive or negative. The fundamental relationship is R2=r2R^2 = r^2. Since R2=0.81R^2 = 0.81, we know that r2=0.81r^2 = 0.81. To find r, we take the square root: r=±0.81=±0.9r = \pm\sqrt{0.81} = \pm 0.9. The problem states that r is negative, so r=−0.9r = -0.9. This makes choice B correct. Let's examine why the other options are wrong. Choice A (r=−0.81r = -0.81) incorrectly assumes that R2R^2 and r have the same numerical value when r is negative. This confuses the relationship between these two statistics. Choice C (r=0.9r = 0.9) correctly calculates the square root but ignores the given information that the correlation is negative. Choice D (r=−0.656r = -0.656) appears to use some incorrect formula, possibly confusing R2R^2 with other statistical measures. Remember that R2R^2 always ranges from 0 to 1 and represents the proportion of variance explained, while r ranges from -1 to +1 and indicates both the strength and direction of the linear relationship. The sign of r tells you the direction of the relationship (negative r means as one variable increases, the other decreases), but R2R^2 only tells you the strength since squaring eliminates the sign.

Question 13

A regression model predicting weight loss (Y) from exercise minutes per day (X) gives Y^=0.5+0.03X\hat{Y} = 0.5 + 0.03X. If the standard deviation of weight loss in the sample is 2.1 pounds and R2=0.25R^2 = 0.25, what is the approximate standard deviation of the residuals?

  1. 1.82 pounds, calculated as sY1−R2s_Y\sqrt{1-R^2} where sYs_Y is the standard deviation of Y (correct answer)
  2. 0.525 pounds, calculated as the standard deviation of Y times R2R^2
  3. 1.05 pounds, calculated as half the standard deviation of Y
  4. 2.1 pounds, which equals the original standard deviation of the dependent variable
  5. 0.75 pounds, calculated as the standard deviation of Y times the square root of R2R^2
Explanation: When you encounter regression problems involving residuals and R2R^2, remember that residuals represent the "leftover" variation after your model explains what it can. The key relationship here connects the total variation in Y to both explained and unexplained components. The standard deviation of residuals measures how much the actual Y values deviate from the predicted Y^\hat{Y} values. This is calculated using the formula sresiduals=sY1−R2s_{residuals} = s_Y\sqrt{1-R^2}, where sYs_Y is the standard deviation of the dependent variable. With sY=2.1s_Y = 2.1 and R2=0.25R^2 = 0.25, we get: sresiduals=2.11−0.25=2.10.75=2.1×0.866=1.82s_{residuals} = 2.1\sqrt{1-0.25} = 2.1\sqrt{0.75} = 2.1 × 0.866 = 1.82 pounds. Choice A correctly applies this fundamental formula. Choice B incorrectly multiplies by R2R^2 instead of 1−R2\sqrt{1-R^2}, which would give you a measure related to explained variation rather than unexplained variation. Choice C arbitrarily divides the standard deviation by 2, which has no theoretical basis in regression analysis. Choice D suggests residual variation equals total variation, which would only be true if the model explained nothing (R2=0R^2 = 0). The intuition is that R2=0.25R^2 = 0.25 means your model explains 25% of the variation, leaving 75% unexplained. The residual standard deviation reflects this remaining 75% of variation. Remember: residual variation = total variation × proportion unexplained\sqrt{\text{proportion unexplained}}. This formula appears frequently in biostatistics, so memorize sresiduals=sY1−R2s_{residuals} = s_Y\sqrt{1-R^2}.

Question 14

A researcher calculates that R2=0.16R^2 = 0.16 for predicting anxiety scores from caffeine intake. A colleague argues this indicates a weak relationship not worth investigating. What is the most appropriate response?

  1. The colleague is correct because R2<0.5R^2 < 0.5 indicates no meaningful linear relationship exists
  2. The relationship may be meaningful; R2=0.16R^2 = 0.16 means caffeine explains 16% of anxiety variance (correct answer)
  3. The colleague is wrong because any positive R2R^2 value indicates a strong relationship
  4. The analysis should be repeated because R2R^2 values this low indicate calculation errors
  5. The relationship is moderate since r=0.16r = 0.16 represents a reasonable correlation coefficient
Explanation: When interpreting R2R^2 values in biostatistics, you need to consider both the statistical meaning and the practical context. R2R^2 represents the proportion of variance in the dependent variable explained by the independent variable(s), expressed as a decimal between 0 and 1. An R2=0.16R^2 = 0.16 means caffeine intake explains 16% of the variance in anxiety scores. While this might seem "small," it's actually meaningful in behavioral research where many factors influence outcomes. Human behavior is complex, so finding that a single variable explains 16% of the variance can be quite substantial and worth investigating further. Option A is incorrect because there's no universal threshold where R2<0.5R^2 < 0.5 indicates "no meaningful relationship." The meaningfulness depends on the research context, sample size, and field of study. In psychology and health research, even smaller R2R^2 values can be practically significant. Option C is wrong because it overstates the case in the opposite direction. Any positive R2R^2 doesn't automatically indicate a "strong" relationship – strength is relative and context-dependent. Option D is incorrect because low R2R^2 values don't suggest calculation errors. They simply reflect that the predictor variable explains a smaller portion of the outcome variance, which is common and valid in many research scenarios. Study tip: Remember that R2R^2 interpretation is field-specific. In biostatistics, don't dismiss relationships based solely on R2R^2 magnitude – consider the research context, clinical significance, and whether the relationship could inform meaningful interventions or understanding.

Question 15

Two regression models are fitted to the same data: Model A has R2=0.72R^2 = 0.72 and Model B has R2=0.68R^2 = 0.68. What can be concluded about the residuals from these models?

  1. Model A residuals have smaller variance than Model B residuals on average (correct answer)
  2. Model A residuals are always smaller in absolute value than Model B residuals
  3. Model B residuals sum to a larger positive value than Model A residuals
  4. The correlation between Model A and Model B residuals equals 0.04
  5. Model A residuals show less bias than Model B residuals in systematic error
Explanation: When you encounter questions about R2R^2 and residuals, remember that R2R^2 measures the proportion of variance in the dependent variable explained by the model. A higher R2R^2 means the model explains more variance, leaving less unexplained variance in the residuals. Since Model A has R2=0.72R^2 = 0.72 and Model B has R2=0.68R^2 = 0.68, Model A explains more of the total variance in the data. This means Model A leaves less unexplained variance, so its residuals will have smaller variance on average compared to Model B's residuals. The relationship is: residual variance = total variance × (1 - R2R^2). Since Model A has higher R2R^2, it has lower residual variance. Option A correctly identifies this fundamental relationship between R2R^2 and residual variance. Option B is wrong because R2R^2 tells us about average patterns, not that every individual residual from Model A is smaller than every residual from Model B. Option C is incorrect because residuals from properly fitted regression models sum to zero (or very close to zero) regardless of R2R^2 value - this is a mathematical property of least squares regression. Option D makes no sense because 0.04 is simply the difference between the R2R^2 values, not a correlation coefficient between residual sets. Study tip: Remember that higher R2R^2 always means smaller residual variance. When comparing models on the same data, the model with higher R2R^2 fits better and produces residuals with less variability around zero.

Question 16

Two variables have a correlation coefficient of r = -0.6. After fitting a regression line, what percentage of the total variance in the dependent variable remains unexplained?

  1. 40% of the variance remains unexplained by the regression model
  2. 60% of the variance remains unexplained by the regression model
  3. 64% of the variance remains unexplained by the regression model (correct answer)
  4. 36% of the variance remains unexplained by the regression model
  5. The percentage cannot be determined without knowing the slope value
Explanation: When you encounter correlation and regression problems, remember that the correlation coefficient (r) and the coefficient of determination (r²) tell different stories about the relationship between variables. The correlation coefficient r = -0.6 tells us there's a moderately strong negative linear relationship between the variables. However, to determine how much variance is explained by the regression model, you need to calculate r² (the coefficient of determination). Here: r² = (-0.6)² = 0.36. This means 36% of the variance in the dependent variable is explained by the regression model. Since 36% of variance is explained, the remaining unexplained variance is: 100% - 36% = 64%. Looking at the answer choices: Choice A (40%) appears to confuse the correlation coefficient with explained variance - this would be correct if r² equaled 0.6, but correlation coefficients must be squared. Choice B (60%) takes the absolute value of the correlation coefficient as unexplained variance, ignoring the squaring step entirely. Choice D (36%) gives you the explained variance rather than the unexplained portion - a common reversal error. Choice C (64%) correctly represents the unexplained variance. Study tip: Always remember the formula: Unexplained variance = 1 - r². The correlation coefficient must be squared first (regardless of whether it's positive or negative), then subtracted from 1 to find the unexplained portion. This relationship between r and r² appears frequently on biostatistics exams.

Question 17

Examine the table of regression results shown. If the research goal is to predict Y with maximum accuracy, which model should be selected and why?

  1. Model 1 because it has the largest slope coefficient indicating strongest relationship
  2. Model 2 because it has the smallest intercept value closest to zero
  3. Model 3 because it has the highest R2R^2 value explaining most variance (correct answer)
  4. Model 2 because its negative slope provides better mathematical balance
  5. Model 1 because its positive intercept ensures all predictions will be realistic
Explanation: R2R^2 measures the proportion of variance explained by the model. Higher R2R^2 indicates better predictive accuracy. Model 3 with R2=0.78R^2 = 0.78 explains the most variance and should provide the most accurate predictions. Choice A incorrectly focuses on slope magnitude. Choice B incorrectly emphasizes intercept size. Choice D provides no valid reasoning. Choice E incorrectly links intercept sign to prediction realism.

Question 18

Based on the scatter plot shown, which statement best describes the residuals for the fitted regression line?

  1. The residuals show a clear curved pattern, suggesting the linear model is appropriate
  2. The residuals are randomly scattered around zero, indicating a good linear fit
  3. The residuals show increasing variance from left to right, violating homoscedasticity (correct answer)
  4. The residuals are all positive, indicating the model consistently overestimates
  5. The residuals show a decreasing trend, suggesting the slope should be steeper
Explanation: The scatter plot shows points that fan out as X increases, creating a funnel shape. This indicates heteroscedasticity (non-constant variance), which violates the assumption of homoscedasticity. Choice A incorrectly suggests a curved pattern supports linearity. Choice B would describe ideal residuals. Choice D mischaracterizes the pattern. Choice E incorrectly relates the pattern to slope steepness.

Question 19

Using the residual plot shown, which conclusion about the regression model is most justified?

  1. The linear model fits well because most residuals are close to the regression line
  2. The model should be rejected because several residuals are negative values
  3. A quadratic model might fit better due to the curved pattern in residuals (correct answer)
  4. The model is perfect because the residuals average to approximately zero
  5. The large residuals indicate the model explains very little variance in the data
Explanation: A curved pattern in residuals (systematic departure from randomness) suggests the linear model misses some curvature in the relationship, indicating a quadratic or other non-linear model might be more appropriate. Choice A misunderstands residual plots. Choice B incorrectly views negative residuals as problematic. Choice D confuses zero mean with good fit. Choice E incorrectly infers R2R^2 from residual magnitude.

Question 20

Based on the scatter plot shown with fitted regression line, which observation has the largest positive residual?

  1. Point A, located in the upper-left region of the plot (correct answer)
  2. Point B, located in the lower-right region of the plot
  3. Point C, located directly on the fitted regression line
  4. Point D, located in the upper-right region of the plot
  5. Point E, located in the lower-left region of the plot
Explanation: A positive residual occurs when the actual Y-value is above the predicted Y-value (above the regression line). The largest positive residual corresponds to the point farthest above the regression line. Point A in the upper-left shows the greatest vertical distance above the line. Point B is below the line (negative residual). Point C is on the line (zero residual). Points D and E are either on or below the line.