Biostatistics Quiz: Interaction Terms
20 questions · exam conditions
0:00
Interaction TermsQuestion 1 of 20

In a study of cognitive test scores, the model is: Score = 85 + 2.5(Education Years) + 8.2(Training) - 0.3(Education Years × Training), where Training = 1 if received special training and 0 otherwise. A researcher wants to know at what education level the training effect becomes negligible (close to zero). What education level satisfies this condition?

Approximately 27 years of education when the training coefficient approaches zero
Approximately 17 years of education when the training coefficient approaches zero
Approximately 8 years of education when the training coefficient approaches zero
The training effect never becomes negligible regardless of education level achieved
← Back to quizzes

Biostatistics Quiz

Biostatistics Quiz: Interaction Terms

Practice Interaction Terms 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 Interaction Terms, 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 cognitive test scores, the model is: Score = 85 + 2.5(Education Years) + 8.2(Training) - 0.3(Education Years × Training), where Training = 1 if received special training and 0 otherwise. A researcher wants to know at what education level the training effect becomes negligible (close to zero). What education level satisfies this condition?

  1. Approximately 27 years of education when the training coefficient approaches zero (correct answer)
  2. Approximately 17 years of education when the training coefficient approaches zero
  3. Approximately 8 years of education when the training coefficient approaches zero
  4. The training effect never becomes negligible regardless of education level achieved
Explanation: The training effect = 8.2 + (-0.3)(Education Years). Setting this equal to 0: 8.2 - 0.3(Education Years) = 0, so Education Years = 8.2/0.3 ≈ 27.3 years. Choice B uses an incorrect calculation (8.2 + 0.3 instead of division). Choice C uses just the interaction coefficient magnitude. Choice D ignores that the negative interaction can eventually overcome the positive main effect.

Question 2

In a nutrition study, weight change (pounds) is modeled as: Weight_Change = -2 + 0.3(Calories) + 4(Exercise_Program) - 0.1(Calories × Exercise_Program), where Calories are in hundreds per day and Exercise_Program = 1 if enrolled, 0 otherwise. At what calorie intake do participants in the exercise program and control group have the same predicted weight change?

  1. At 1000 calories per day, both groups have identical weight changes
  2. At 2000 calories per day, both groups have identical weight changes
  3. At 4000 calories per day, both groups have identical weight changes (correct answer)
  4. The groups never have identical weight changes at any calorie level
  5. At 1500 calories per day, both groups have identical weight changes
Explanation: When you encounter regression models with interaction terms, you're looking at how the effect of one variable changes depending on the level of another variable. Here, the interaction term -0.1(\text{Calories} \times \text{Exercise_Program}) means the effect of calories differs between the exercise and control groups. To find where both groups have identical weight changes, set up equations for each group. For the control group (Exercise_Program = 0): \text{Weight_Change} = -2 + 0.3(\text{Calories}). For the exercise group (Exercise_Program = 1): \text{Weight_Change} = -2 + 0.3(\text{Calories}) + 4 - 0.1(\text{Calories}) = 2 + 0.2(\text{Calories}). Setting these equal: −2+0.3(Calories)=2+0.2(Calories)-2 + 0.3(\text{Calories}) = 2 + 0.2(\text{Calories}). Solving: 0.1(Calories)=40.1(\text{Calories}) = 4, so Calories=40\text{Calories} = 40 (in hundreds per day), which equals 4000 calories per day. Answer A (1000 calories) gives different weight changes: control group = +1 pound, exercise group = +4 pounds. Answer B (2000 calories) also differs: control group = +4 pounds, exercise group = +6 pounds. Answer D is incorrect because we just demonstrated the groups do have identical changes at one specific calorie level. Answer C correctly identifies 4000 calories as the intersection point. Remember: interaction terms create different slopes for each group, and finding where outcomes are identical means solving for where these different linear relationships intersect. Always set up separate equations for each group when dealing with categorical variables in regression.

Question 3

In a clinical trial, the relationship between drug dose (mg) and symptom improvement is modeled as: Improvement = β₀ + β₁(Dose) + β₂(Treatment_Type) + β₃(Dose × Treatment_Type), where Treatment_Type = 0 for monotherapy and 1 for combination therapy. If β₁ = 0.5, β₂ = 8, and β₃ = -0.3, which statement correctly describes the interaction effect?

  1. Combination therapy reduces the dose-response slope by 0.3 units per mg compared to monotherapy (correct answer)
  2. Combination therapy increases the dose-response slope by 0.3 units per mg compared to monotherapy
  3. The interaction indicates that combination therapy is always 8 points better than monotherapy
  4. Monotherapy has a steeper dose-response relationship than combination therapy by 0.8 units per mg
  5. The interaction effect is only significant when the dose exceeds 8 mg in combination therapy
Explanation: When you encounter regression models with interaction terms, focus on how the interaction coefficient modifies the relationship between variables. In this dose-response model, the interaction term β3(Dose×TreatmentType)β₃(Dose × Treatment_Type) tells you how treatment type changes the dose-response slope. To find the dose-response slope for each treatment type, take the derivative with respect to dose. For monotherapy (Treatment_Type = 0): slope = β1=0.5β₁ = 0.5. For combination therapy (Treatment_Type = 1): slope = β1+β3=0.5+(−0.3)=0.2β₁ + β₃ = 0.5 + (-0.3) = 0.2. The interaction effect is the difference in slopes: combination therapy's slope is 0.3 units per mg lower than monotherapy's slope. Answer A correctly states that combination therapy reduces the dose-response slope by 0.3 units per mg compared to monotherapy. Answer B is backwards—it claims combination therapy increases the slope, but β3=−0.3β₃ = -0.3 indicates a decrease. Answer C misinterprets β2=8β₂ = 8, which represents the baseline difference between treatments at dose = 0, not a constant advantage regardless of dose. Answer D incorrectly calculates the difference as 0.8 units—this likely comes from adding the absolute values of β1β₁ and β3β₃ rather than properly computing the slope difference. Study tip: For interaction terms, always calculate the slope or effect for each group separately, then compare. The interaction coefficient shows how much one group's relationship differs from the reference group's relationship—pay attention to the sign to determine direction.

Question 4

A regression model predicts hospital length of stay (days) using patient age and surgery type: Stay = 2 + 0.1(Age) + 3(Surgery) + 0.05(Age × Surgery), where Surgery = 0 for outpatient, 1 for inpatient. For a 60-year-old patient, what is the difference in predicted length of stay between inpatient and outpatient surgery?

  1. The difference is exactly 3 days longer for inpatient surgery
  2. The difference is 5 days longer for inpatient surgery
  3. The difference is 6 days longer for inpatient surgery (correct answer)
  4. The difference is 8 days longer for inpatient surgery
  5. The difference is 3.05 days longer for inpatient surgery
Explanation: When you encounter regression models with interaction terms, you need to calculate predictions for both scenarios and find their difference, paying careful attention to how the interaction term changes. For this model: Stay=2+0.1(Age)+3(Surgery)+0.05(Age×Surgery)\text{Stay} = 2 + 0.1(\text{Age}) + 3(\text{Surgery}) + 0.05(\text{Age} \times \text{Surgery}) For a 60-year-old with outpatient surgery (Surgery = 0): Stay=2+0.1(60)+3(0)+0.05(60×0)=2+6+0+0=8 days\text{Stay} = 2 + 0.1(60) + 3(0) + 0.05(60 \times 0) = 2 + 6 + 0 + 0 = 8 \text{ days} For a 60-year-old with inpatient surgery (Surgery = 1): Stay=2+0.1(60)+3(1)+0.05(60×1)=2+6+3+3=14 days\text{Stay} = 2 + 0.1(60) + 3(1) + 0.05(60 \times 1) = 2 + 6 + 3 + 3 = 14 \text{ days} The difference is 14 - 8 = 6 days longer for inpatient surgery, making C correct. A is wrong because it ignores the interaction term entirely, only considering the main effect of surgery (3 days). B incorrectly adds just the age coefficient to the surgery effect (3 + 0.1×60 = 8, but this isn't how interactions work). D doubles the interaction effect or makes some other calculation error. The key insight is that interaction terms modify the relationship between variables. Here, the effect of surgery type depends on age - older patients show a bigger difference between surgery types than younger patients would. Always calculate both complete predictions when dealing with interaction terms, rather than trying to shortcut the math.

Question 5

A researcher models salary (thousands) as: Salary = 40 + 2(Experience) + 8(Advanced_Degree) + 1.5(Experience × Advanced_Degree), where Experience is in years and Advanced_Degree = 1 if present, 0 otherwise. For someone with an advanced degree, what is the marginal effect of one additional year of experience?

  1. Each additional year increases salary by $2,000 for those with advanced degrees
  2. Each additional year increases salary by $3,500 for those with advanced degrees (correct answer)
  3. Each additional year increases salary by $1,500 for those with advanced degrees
  4. Each additional year increases salary by $8,000 for those with advanced degrees
  5. Each additional year increases salary by $9,500 for those with advanced degrees
Explanation: When you encounter regression models with interaction terms, the key is understanding how the marginal effect of one variable depends on the value of another variable. The interaction term (Experience × Advanced_Degree) means the effect of experience differs based on whether someone has an advanced degree. To find the marginal effect of experience for someone with an advanced degree, you take the partial derivative of salary with respect to experience, setting Advanced_Degree = 1. From the equation Salary = 40 + 2(Experience) + 8(Advanced_Degree) + 1.5(Experience × Advanced_Degree), the marginal effect is: ∂Salary∂Experience=2+1.5×AdvancedDegree\frac{\partial Salary}{\partial Experience} = 2 + 1.5 \times Advanced_Degree For someone with an advanced degree (Advanced_Degree = 1): ∂Salary∂Experience=2+1.5(1)=3.5\frac{\partial Salary}{\partial Experience} = 2 + 1.5(1) = 3.5 Since salary is measured in thousands, each additional year increases salary by $3,500. Answer A (2,000)representsonlythebaseeffectofexperience,ignoringtheinteractiontermentirely.Thiswouldbecorrectforsomeonewithoutanadvanceddegree.AnswerC(2,000) represents only the base effect of experience, ignoring the interaction term entirely. This would be correct for someone without an advanced degree. Answer C (1,500) captures only the interaction effect, missing the base effect of experience. Answer D ($8,000) confuses the coefficient for having an advanced degree with the marginal effect of experience. Study tip: In interaction models, always remember that marginal effects are conditional. When calculating the effect of one variable, include both its direct coefficient and any interaction terms where the other variable takes its specified value. Practice identifying which variables are "turned on" in your scenario.

Question 6

A psychological study models stress level as: Stress = 50 + 3(Workload) - 10(Social_Support) - 0.5(Workload × Social_Support), where Workload and Social_Support are measured on 0-10 scales. If Social_Support increases from 2 to 6 while Workload remains at 8, what is the predicted change in stress level?

  1. Stress decreases by 40 points due to increased social support
  2. Stress decreases by 24 points accounting for interaction effects
  3. Stress decreases by 56 points from combined main and interaction effects (correct answer)
  4. Stress decreases by 16 points from the interaction effect alone
  5. Stress decreases by 32 points when all model components are considered
Explanation: When you encounter regression models with interaction terms, you must calculate the total effect by considering both main effects and how variables modify each other's impact. To find the predicted change in stress, calculate the stress level at both Social_Support values while keeping Workload constant at 8. At Social_Support = 2: Stress=50+3(8)−10(2)−0.5(8×2)=50+24−20−8=46Stress = 50 + 3(8) - 10(2) - 0.5(8 \times 2) = 50 + 24 - 20 - 8 = 46 At Social_Support = 6: Stress=50+3(8)−10(6)−0.5(8×6)=50+24−60−24=−10Stress = 50 + 3(8) - 10(6) - 0.5(8 \times 6) = 50 + 24 - 60 - 24 = -10 The change in stress is: −10−46=−56-10 - 46 = -56 points, meaning stress decreases by 56 points. Answer A (40 points) only considers the main effect of Social_Support: −10×(6−2)=−40-10 \times (6-2) = -40, ignoring the interaction term entirely. Answer B (24 points) appears to miscalculate by considering only part of the interaction effect. Answer D (16 points) isolates just the interaction effect: −0.5×8×(6−2)=−16-0.5 \times 8 \times (6-2) = -16, but misses that you need the total change including main effects. The interaction term −0.5(Workload×SocialSupport)-0.5(Workload \times Social_Support) means that social support becomes more effective at reducing stress when workload is higher. At Workload = 8, each unit increase in social support provides additional stress reduction beyond its main effect. Study tip: With interaction terms, always calculate the full model at both conditions rather than trying to separate main and interaction effects—the total change is what matters for interpretation.

Question 7

A study of learning outcomes uses the model: Test_Score = 70 + 1.5(Study_Hours) + 12(Tutor) - 0.4(Study_Hours × Tutor), where Tutor = 1 if tutoring received, 0 otherwise. For students who receive tutoring, at how many study hours does the predicted test score equal that of a non-tutored student studying the same amount?

  1. This scenario is impossible since tutored students always score higher (correct answer)
  2. At 30 study hours, both groups achieve the same predicted score
  3. At 10.9 study hours, both groups achieve the same predicted score
  4. At 20 study hours, both groups achieve the same predicted score
  5. At 8 study hours, both groups achieve the same predicted score
Explanation: When you encounter regression models with interaction terms, you need to carefully analyze how the coefficients work together to understand the relationship between variables. Let's set up equations for both groups. For non-tutored students (Tutor = 0): Score=70+1.5(Study Hours)\text{Score} = 70 + 1.5(\text{Study Hours}) For tutored students (Tutor = 1): Score=70+1.5(Study Hours)+12−0.4(Study Hours)\text{Score} = 70 + 1.5(\text{Study Hours}) + 12 - 0.4(\text{Study Hours}) =82+1.1(Study Hours)= 82 + 1.1(\text{Study Hours}) The key insight is comparing the slopes: non-tutored students gain 1.5 points per study hour, while tutored students only gain 1.1 points per study hour. Tutored students start with a 12-point advantage (82 vs 70 intercept), but their gains diminish more quickly. To find when scores are equal, set the equations equal: 70+1.5h=82+1.1h70 + 1.5h = 82 + 1.1h 0.4h=120.4h = 12 h=30h = 30 However, this calculation shows that tutored students would need to study 30 hours to lose their advantage completely. Since tutored students start ahead and only gradually lose ground, they never actually fall behind or even tie with non-tutored students—they always maintain some advantage. Answer A is correct because tutored students always score higher, regardless of study hours. Answers B, C, and D all suggest specific hours where scores equalize, but this misinterprets the interaction effect. While the gap narrows over time, tutored students maintain their edge. Study tip: With interaction terms, always calculate the actual equations for each group separately, then examine both intercepts and slopes to understand the full relationship.

Question 8

A fitness study models calorie burn as: Calories = 200 + 15(Duration) + 50(High_Intensity) + 5(Duration × High_Intensity), where Duration is in hours and High_Intensity = 1 if high-intensity workout, 0 for moderate. For a 1.5-hour workout, what is the predicted difference in calorie burn between high-intensity and moderate-intensity exercise?

  1. High-intensity burns exactly 50 more calories than moderate-intensity
  2. High-intensity burns 57.5 more calories than moderate-intensity exercise (correct answer)
  3. High-intensity burns 65 more calories than moderate-intensity exercise
  4. High-intensity burns 42.5 more calories than moderate-intensity exercise
  5. High-intensity burns 7.5 more calories than moderate-intensity exercise
Explanation: When you encounter a regression model with interaction terms, you need to calculate the total effect by considering both the main effect and the interaction effect together. This model includes an interaction term (Duration × High_Intensity), which means the effect of intensity depends on workout duration. To find the difference between high-intensity and moderate-intensity workouts at 1.5 hours, calculate each scenario separately: Moderate-intensity (High_Intensity = 0): Calories=200+15(1.5)+50(0)+5(1.5×0)=200+22.5=222.5Calories = 200 + 15(1.5) + 50(0) + 5(1.5 × 0) = 200 + 22.5 = 222.5 High-intensity (High_Intensity = 1): Calories=200+15(1.5)+50(1)+5(1.5×1)=200+22.5+50+7.5=280Calories = 200 + 15(1.5) + 50(1) + 5(1.5 × 1) = 200 + 22.5 + 50 + 7.5 = 280 The difference is 280−222.5=57.5280 - 222.5 = 57.5 calories, confirming answer B. Answer A (50 calories) represents only the main effect coefficient for High_Intensity, ignoring the interaction term completely. Answer C (65 calories) incorrectly adds the full interaction coefficient (5) to the main effect rather than multiplying it by the duration. Answer D (42.5 calories) appears to subtract the interaction effect instead of adding it, suggesting a calculation error. Study tip: With interaction terms, always substitute the actual values for all variables rather than trying to estimate effects using coefficients alone. The interaction term changes the relationship between variables depending on their values, so you must calculate the full equation for meaningful comparisons.

Question 9

A sleep quality study models: Sleep_Quality = 50 + 2(Exercise_Hours) + 10(Meditation) - 0.5(Exercise_Hours × Meditation), where Meditation = 1 if practiced, 0 otherwise. At what level of exercise hours do people who meditate and those who don't achieve the same predicted sleep quality?

  1. At 20 hours of exercise per week, both groups have equal sleep quality (correct answer)
  2. At 15 hours of exercise per week, both groups have equal sleep quality
  3. At 5 hours of exercise per week, both groups have equal sleep quality
  4. The groups never achieve equal sleep quality at any exercise level
  5. At 10 hours of exercise per week, both groups have equal sleep quality
Explanation: When you encounter regression models with interaction terms, you're looking at how the effect of one variable changes depending on the level of another variable. Here, the interaction term (Exercise_Hours × Meditation) means the slope for exercise differs between meditators and non-meditators. To find where both groups achieve equal sleep quality, you need to set up equations for each group and solve where they're equal. For non-meditators (Meditation = 0): Sleep_Quality = 50 + 2(Exercise_Hours). For meditators (Meditation = 1): Sleep_Quality = 50 + 2(Exercise_Hours) + 10 - 0.5(Exercise_Hours), which simplifies to Sleep_Quality = 60 + 1.5(Exercise_Hours). Setting these equal: 50+2x=60+1.5x50 + 2x = 60 + 1.5x, where x = exercise hours. Solving: 2x−1.5x=60−502x - 1.5x = 60 - 50, so 0.5x=100.5x = 10, therefore x=20x = 20 hours. Answer A correctly identifies 20 hours as the intersection point. Answer B (15 hours) would give different sleep quality scores for each group when you substitute back into the equations. Answer C (5 hours) similarly produces unequal outcomes when calculated. Answer D incorrectly assumes the lines never intersect, but since they have different slopes (2 vs 1.5), they must intersect exactly once. Study tip: For interaction term problems, always write out the complete equation for each group separately, then set them equal to find intersection points. The interaction term creates different slopes, making intersection calculations straightforward algebra.

Question 10

In an environmental study, air quality index is predicted by: AQI = 100 - 5(Wind_Speed) - 8(Rain) + 2(Wind_Speed × Rain), where Wind_Speed is in mph and Rain = 1 if present, 0 if absent. What happens to the effect of wind speed when it rains compared to when it doesn't rain?

  1. Wind speed becomes less effective at reducing AQI when it rains (correct answer)
  2. Wind speed becomes more effective at reducing AQI when it rains
  3. Wind speed has no effect on AQI when it rains
  4. Wind speed increases AQI when it rains instead of decreasing it
  5. The effect of wind speed remains identical regardless of rain conditions
Explanation: When you encounter regression equations with interaction terms, focus on how coefficients change under different conditions. This question tests your ability to interpret interaction effects in multiple regression. To find wind speed's effect, look at its coefficient under each rain condition. When there's no rain (Rain = 0), the equation becomes: AQI = 100 - 5(Wind_Speed). Here, wind speed's coefficient is -5, meaning each mph of wind speed reduces AQI by 5 points. When it rains (Rain = 1), substitute into the full equation: AQI = 100 - 5(Wind_Speed) - 8(1) + 2(Wind_Speed × 1) = 92 - 3(Wind_Speed). Now wind speed's effective coefficient is -3 (combining the -5 and +2 terms). Since |-3| < |-5|, wind speed is less effective at reducing AQI when it rains, making A correct. B is wrong because wind speed's effect weakens (from -5 to -3), not strengthens. C is incorrect because wind speed still has an effect (-3 coefficient), just a reduced one. D misinterprets the math—the coefficient remains negative (-3), so wind speed still decreases AQI rather than increasing it. The key insight is that the interaction term (+2) partially counteracts wind speed's beneficial effect during rain, possibly because rain reduces wind's ability to disperse pollutants. Study tip: For interaction terms, always calculate the effective coefficient by combining the main effect with the interaction effect. The sign and magnitude of this combined coefficient tell the complete story.

Question 11

An educational technology study uses: Performance = 75 + 3(Practice_Time) + 15(Technology_Use) - 1(Practice_Time × Technology_Use), where Practice_Time is in hours and Technology_Use = 1 if technology is used, 0 otherwise. For students using technology, what is the effective slope of practice time on performance?

  1. The effective slope is 2 points per hour for technology users (correct answer)
  2. The effective slope is 3 points per hour for technology users
  3. The effective slope is 4 points per hour for technology users
  4. The effective slope is 15 points per hour for technology users
  5. The effective slope is 1 point per hour for technology users
Explanation: When you encounter a regression equation with an interaction term, you need to understand how the interaction changes the relationship between variables for different groups. The interaction term modifies the main effect depending on the value of the other variable. To find the effective slope of practice time for technology users, substitute Technology_Use = 1 into the equation. The performance equation becomes: Performance = 75 + 3(Practice_Time) + 15(1) - 1(Practice_Time × 1) = 75 + 3(Practice_Time) + 15 - 1(Practice_Time). Simplifying the practice time terms: 3(Practice_Time) - 1(Practice_Time) = 2(Practice_Time). So for technology users, Performance = 90 + 2(Practice_Time), giving an effective slope of 2 points per hour. Answer A correctly identifies this effective slope of 2 points per hour. Answer B (3 points per hour) represents the main effect coefficient for practice time, ignoring the interaction term's modification. Answer C (4 points per hour) incorrectly adds the interaction coefficient instead of subtracting it (3 + 1 = 4). Answer D (15 points per hour) confuses the technology main effect with the practice time slope. Remember that interaction terms always modify main effects. When calculating effective slopes in the presence of interactions, substitute the specific value of the moderating variable (here, Technology_Use = 1) and combine all terms involving your variable of interest. The interaction coefficient gets multiplied by both variables, so it affects the slope when the other variable equals 1.

Question 12

In a regression model predicting blood pressure (Y) from age (X₁) and smoking status (X₂, where 0 = non-smoker, 1 = smoker), the interaction term X₁X₂ has a coefficient of -0.8. If the main effect of age is 1.2 mmHg per year, what is the effective slope of age for smokers?

  1. The effective slope is 0.4 mmHg per year for smokers (correct answer)
  2. The effective slope is 1.2 mmHg per year for smokers
  3. The effective slope is 2.0 mmHg per year for smokers
  4. The effective slope cannot be determined without the intercept term
  5. The effective slope is -0.8 mmHg per year for smokers
Explanation: When you encounter interaction terms in regression, you're dealing with how the effect of one variable changes depending on the level of another variable. The interaction term modifies the main effect for different groups. Your regression model is: Y=β0+β1X1+β2X2+β3X1X2Y = \beta_0 + \beta_1X_1 + \beta_2X_2 + \beta_3X_1X_2 For smokers (X₂ = 1), the equation becomes: Y=β0+β1X1+β2(1)+β3X1(1)=β0+(β1+β3)X1+β2Y = \beta_0 + \beta_1X_1 + \beta_2(1) + \beta_3X_1(1) = \beta_0 + (\beta_1 + \beta_3)X_1 + \beta_2 The effective slope of age for smokers is the coefficient of X₁, which is (β1+β3)=1.2+(−0.8)=0.4(\beta_1 + \beta_3) = 1.2 + (-0.8) = 0.4 mmHg per year. This means A is correct — the effective slope is 0.4 mmHg per year for smokers. B is wrong because it ignores the interaction term entirely, giving only the main effect of age (1.2). This would be the slope for non-smokers, not smokers. C is wrong because it incorrectly adds the interaction coefficient as positive (1.2 + 0.8 = 2.0), misunderstanding that the interaction modifies the main effect by -0.8, not +0.8. D is wrong because the intercept term doesn't affect the slope calculation. Slopes represent the rate of change, which depends only on the coefficients of the variable terms, not the intercept. Study tip: In interaction models, always substitute the group's value into the equation to find the effective coefficient. For categorical variables coded 0/1, the group coded as 1 gets both the main effect plus the interaction effect.

Question 13

In a medical device study, the model for measurement accuracy is: Accuracy = 85 + 2(Calibration) + 5(Technician_Training) + 0.8(Calibration × Technician_Training), where both predictors are binary (0 = absent, 1 = present). What is the predicted accuracy when both calibration and technician training are present compared to when both are absent?

  1. Accuracy improves by exactly 7 percentage points with both factors present
  2. Accuracy improves by 7.8 percentage points with both factors present (correct answer)
  3. Accuracy improves by 6.2 percentage points with both factors present
  4. Accuracy improves by 12.8 percentage points with both factors present
  5. Accuracy improves by 14.6 percentage points with both factors present
Explanation: When you encounter a regression model with an interaction term, you need to calculate the combined effect of all relevant terms, not just add the main effects. This model includes both main effects and an interaction between calibration and technician training. Let's calculate the predicted accuracy for both scenarios. When both factors are absent (Calibration = 0, Technician_Training = 0): Accuracy=85+2(0)+5(0)+0.8(0×0)=85Accuracy = 85 + 2(0) + 5(0) + 0.8(0 \times 0) = 85 When both factors are present (Calibration = 1, Technician_Training = 1): Accuracy=85+2(1)+5(1)+0.8(1×1)=85+2+5+0.8=92.8Accuracy = 85 + 2(1) + 5(1) + 0.8(1 \times 1) = 85 + 2 + 5 + 0.8 = 92.8 The improvement is 92.8 - 85 = 7.8 percentage points, making B correct. Option A (7 points) represents the classic mistake of only adding the main effects (2 + 5 = 7) while ignoring the interaction term of 0.8. Option C (6.2 points) might result from incorrectly subtracting rather than adding the interaction term (2 + 5 - 0.8 = 6.2). Option D (12.8 points) appears to double-count something, perhaps adding all terms twice or misunderstanding how interaction terms work. The key insight is that interaction terms capture synergistic effects—when both factors are present simultaneously, they provide an additional boost beyond their individual contributions. Always remember to include interaction terms in your calculations when the interacting variables are both "active" in your scenario.

Question 14

In a pharmaceutical study, drug effectiveness is modeled as: Effectiveness = 20 + 5(Dose) + 12(Patient_Age_Group) + 2(Dose × Patient_Age_Group), where Dose is in mg and Patient_Age_Group = 1 for elderly, 0 for younger adults. For elderly patients, if the dose increases from 10mg to 15mg, what is the predicted change in effectiveness?

  1. Effectiveness increases by 25 points for elderly patients
  2. Effectiveness increases by 35 points for elderly patients (correct answer)
  3. Effectiveness increases by 15 points for elderly patients
  4. Effectiveness increases by 45 points for elderly patients
  5. Effectiveness increases by 10 points for elderly patients
Explanation: When you encounter regression models with interaction terms in biostatistics, you're dealing with situations where the effect of one variable depends on the level of another variable. Here, the interaction term 2(\text{Dose} \times \text{Patient_Age_Group}) means that dose effects differ between age groups. For elderly patients, Patient_Age_Group = 1, so the effectiveness equation becomes: Effectiveness=20+5(Dose)+12(1)+2(Dose×1)=32+7(Dose)\text{Effectiveness} = 20 + 5(\text{Dose}) + 12(1) + 2(\text{Dose} \times 1) = 32 + 7(\text{Dose}) When dose increases from 10mg to 15mg (a 5mg increase), the change in effectiveness is: 7×5=35 points7 \times 5 = 35 \text{ points} The coefficient 7 represents the total effect of dose for elderly patients: the main effect (5) plus the interaction effect (2). Answer A (25 points) incorrectly uses only the main dose effect (5) multiplied by the dose change, ignoring the interaction term entirely. Answer C (15 points) appears to confuse the dose change (5mg) with some other calculation. Answer D (45 points) might result from incorrectly adding all coefficients or misapplying the interaction term. The correct answer is B: effectiveness increases by 35 points for elderly patients. Study tip: With interaction terms, always substitute the specific group values first to get the simplified equation for that subgroup. This prevents errors and makes the calculation straightforward. Remember that interaction terms modify the main effects, so you must account for both when calculating changes within specific groups.

Question 15

A pharmaceutical study models drug concentration with: Concentration = 12.4 + 3.1(Dose) + 5.7(Route) + 2.2(Dose × Route), where Route = 1 for intravenous and 0 for oral administration. If the interaction term p-value = 0.03 and the model R² increases from 0.61 to 0.68 when the interaction is added, what is the most appropriate interpretation?

  1. The interaction explains 68% of the variance in drug concentration levels across administration routes
  2. Intravenous administration always produces higher concentrations than oral regardless of the dose given
  3. The dose-concentration relationship differs significantly between intravenous and oral administration methods (correct answer)
  4. The interaction term alone accounts for 7% improvement in the model's predictive accuracy
Explanation: A significant interaction (p = 0.03) indicates that the relationship between dose and concentration depends on the route of administration - the slopes differ between IV and oral routes. Choice A misinterprets R² as applying only to the interaction. Choice B describes a main effect without considering dose dependency. Choice D incorrectly attributes the R² change solely to predictive accuracy rather than variance explained.

Question 16

A nutrition study models weight change using: Weight Change = -2.1 + 0.4(Calories) + 3.2(Exercise) - 0.1(Calories × Exercise), where Calories is in hundreds per day above baseline and Exercise = 1 for high-intensity program, 0 for moderate-intensity. The interaction term has p = 0.08. What conclusion is most appropriate regarding the calorie-weight relationship?

  1. High-intensity exercise reduces the positive effect of additional calories on weight change significantly
  2. Additional calories have no meaningful effect on weight change in either exercise program group
  3. Moderate-intensity exercise enhances the relationship between calories and weight change more than high-intensity
  4. The calorie effect varies between exercise programs, but this difference is not statistically significant (correct answer)
Explanation: When analyzing regression models with interaction terms, you need to understand how the interaction coefficient modifies the relationship between variables across different groups, and crucially, whether this modification is statistically significant. In this model, the interaction term (Calories × Exercise) has a coefficient of -0.1 with p = 0.08. This tells you two important things: First, the effect of calories on weight change does differ between exercise programs - for moderate-intensity (Exercise = 0), each hundred additional calories increases weight change by 0.4 units, while for high-intensity (Exercise = 1), the effect is 0.4 - 0.1 = 0.3 units. Second, with p = 0.08 > 0.05, this difference is not statistically significant at the conventional alpha level. Answer D correctly captures both aspects: there is a difference in calorie effects between groups, but it's not statistically significant. Answer A incorrectly concludes the reduction is "significant" when p = 0.08 indicates it's not statistically significant. Answer B misses that calories do have meaningful effects in both groups (0.4 and 0.3 units respectively) - the interaction doesn't eliminate the calorie effect entirely. Answer C gets the direction wrong: high-intensity exercise actually reduces the calorie effect compared to moderate-intensity, not the reverse. Study tip: When interpreting interaction terms, always check both the magnitude of the effect AND its statistical significance. A p-value above 0.05 means you cannot conclude the interaction is statistically meaningful, even if the coefficient suggests a numerical difference exists.

Question 17

A sleep study uses the model: Sleep Quality = 6.8 + 0.3(Sleep Hours) + 1.5(Medication) + 0.2(Sleep Hours × Medication), where Medication = 1 for sleep aid users and 0 for non-users. Based on this model, for every additional hour of sleep, what is the predicted improvement in sleep quality for medication users compared to non-users?

  1. Sleep quality improves by 0.3 points per hour more for medication users than non-users
  2. Sleep quality improves by 1.5 points per hour more for medication users than non-users
  3. Sleep quality improves by 0.5 points per hour more for medication users than non-users
  4. Sleep quality improves by 0.2 points per hour more for medication users than non-users (correct answer)
Explanation: When you encounter a regression model with an interaction term, you're looking at how the effect of one variable changes depending on the level of another variable. The interaction coefficient tells you the difference in slopes between groups. To find how the effect of sleep hours differs between medication users and non-users, you need to compare the slopes for each group. For non-users (Medication = 0), the slope for sleep hours is simply the coefficient of Sleep Hours: 0.3. For medication users (Medication = 1), the slope becomes the coefficient of Sleep Hours plus the interaction term: 0.3 + 0.2 = 0.5. The difference between these slopes is 0.5 - 0.3 = 0.2 points per hour, making D correct. Choice A incorrectly uses just the main effect coefficient (0.3) as if it represents the difference between groups, when it actually represents the slope for non-users only. Choice B mistakenly uses the medication main effect (1.5), which represents the baseline difference between groups when sleep hours = 0, not the difference in slopes. Choice C calculates the slope for medication users (0.5) but presents it as the difference between groups rather than recognizing that this is the absolute effect for medication users. Remember: in regression models with interactions, the interaction coefficient directly tells you how much more (or less) one variable's effect is in the presence of the other variable. Don't get distracted by the main effects when the question asks about differences in slopes.

Question 18

A regression model predicting hospital length of stay includes: Stay = β₀ + β₁(Age) + β₂(Surgery Type) + β₃(Age × Surgery Type), where Surgery Type = 1 for major surgery and 0 for minor surgery. If β₁ = 0.12, β₃ = 0.08, and both are significant, what happens to the age effect when comparing surgery types?

  1. Age increases stay by 0.12 days per year for minor surgery and 0.08 days per year for major surgery
  2. Age increases stay by 0.12 days per year for minor surgery and 0.20 days per year for major surgery (correct answer)
  3. Age increases stay by 0.08 days per year for minor surgery and 0.12 days per year for major surgery
  4. Age increases stay by 0.20 days per year for minor surgery and 0.12 days per year for major surgery
Explanation: For minor surgery (Surgery Type = 0): age effect = β₁ = 0.12 days per year. For major surgery (Surgery Type = 1): age effect = β₁ + β₃ = 0.12 + 0.08 = 0.20 days per year. The interaction term adds to the main age effect for major surgery patients. Choice A incorrectly assigns β₃ as the major surgery effect. Choice C reverses the surgery types. Choice D incorrectly adds the interaction to minor surgery.

Question 19

In a study of blood pressure medication effectiveness, the model BP Reduction = β₀ + β₁(Dose) + β₂(Gender) + β₃(Dose × Gender) shows β₁ = 3.2, β₂ = -1.8, and β₃ = -0.7, where Gender = 1 for females and 0 for males. If the interaction term is statistically significant (p < 0.05), what does this indicate about the dose-response relationship?

  1. Males and females have different baseline blood pressures regardless of medication dose administered
  2. The medication dose effect is 0.7 units stronger in males compared to females per unit increase (correct answer)
  3. Females require higher doses than males to achieve the same blood pressure reduction outcomes
  4. The relationship between dose and blood pressure varies significantly between gender groups studied
Explanation: The interaction coefficient β₃ = -0.7 means that for each unit increase in dose, the effect is 0.7 units less for females than males. Since males have slope 3.2 and females have slope 3.2 + (-0.7) = 2.5, the dose effect is stronger in males by 0.7 units per unit dose. Choice A describes a main effect, not interaction. Choice C incorrectly interprets the direction. Choice D is too vague about what 'varies' means.

Question 20

A clinical trial examines pain reduction scores using the model: Pain Reduction = 15.2 + 4.1(Treatment) - 2.3(Severity) + 1.8(Treatment × Severity), where Treatment = 1 for new drug and 0 for placebo, and Severity ranges from 1-5. For a patient with severity score 4, what is the difference in predicted pain reduction between the new drug and placebo?

  1. The new drug provides 4.1 points more pain reduction than placebo for this patient
  2. The new drug provides 7.2 points more pain reduction than placebo for this patient
  3. The new drug provides 11.3 points more pain reduction than placebo for this patient (correct answer)
  4. The new drug provides 1.8 points more pain reduction than placebo for this patient
Explanation: For new drug (Treatment = 1): Pain Reduction = 15.2 + 4.1(1) - 2.3(4) + 1.8(1×4) = 15.2 + 4.1 - 9.2 + 7.2 = 17.3. For placebo (Treatment = 0): Pain Reduction = 15.2 + 4.1(0) - 2.3(4) + 1.8(0×4) = 15.2 - 9.2 = 6.0. Difference = 17.3 - 6.0 = 11.3 points. Choice A uses only the main treatment effect. Choice B omits the severity main effect. Choice D uses only the interaction coefficient.