Biostatistics Quiz: Missing Data Concepts
20 questions · exam conditions
0:00
Missing Data ConceptsQuestion 1 of 20

In a survey about financial stress, participants are asked about both their current debt levels and their confidence in financial security. Missing data analysis reveals that participants with missing debt data have significantly lower financial confidence scores, while participants with missing confidence data have significantly higher reported debt levels. What analytical approach would be most appropriate given this missing data pattern?

Complete case analysis because the systematic patterns indicate potential bias in any imputation
Multiple imputation using each variable to predict missingness in the other under MAR assumptions
Sensitivity analysis comparing MAR and MNAR approaches due to potential self-selection patterns
Separate analysis of each variable to avoid complications from cross-variable missing patterns
Pattern mixture modeling to explicitly account for different missing data subgroups
← Back to quizzes

Biostatistics Quiz

Biostatistics Quiz: Missing Data Concepts

Practice Missing Data Concepts 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 Missing Data Concepts, 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 survey about financial stress, participants are asked about both their current debt levels and their confidence in financial security. Missing data analysis reveals that participants with missing debt data have significantly lower financial confidence scores, while participants with missing confidence data have significantly higher reported debt levels. What analytical approach would be most appropriate given this missing data pattern?

  1. Complete case analysis because the systematic patterns indicate potential bias in any imputation
  2. Multiple imputation using each variable to predict missingness in the other under MAR assumptions (correct answer)
  3. Sensitivity analysis comparing MAR and MNAR approaches due to potential self-selection patterns
  4. Separate analysis of each variable to avoid complications from cross-variable missing patterns
  5. Pattern mixture modeling to explicitly account for different missing data subgroups
Explanation: When analyzing missing data patterns, you need to identify the mechanism behind the missingness and choose methods that can handle the specific relationships you observe. This scenario describes a clear pattern where missingness in one variable is systematically related to observed values in another variable. The pattern you're seeing—where people with missing debt data have lower confidence scores, and those with missing confidence data have higher debt levels—suggests the data are Missing at Random (MAR). Under MAR, the probability of missingness depends on observed variables, which is exactly what's happening here. Multiple imputation is specifically designed for MAR scenarios and can use the observed values of each variable to predict missing values in the other, leveraging the systematic relationships you've identified. Option A is wrong because systematic patterns don't automatically disqualify imputation—they actually provide valuable information for MAR-based methods. Option C incorrectly assumes this might be Missing Not at Random (MNAR), but the clear relationship between missingness and observed variables points to MAR, not MNAR where missingness would depend on unobserved values. Option D wastes the cross-variable information that could improve your analysis and fails to address the missing data properly. Multiple imputation will create several complete datasets using the observed relationships, analyze each one, and pool the results to account for imputation uncertainty. This approach maintains the sample size and uses all available information efficiently. Study tip: When you see missing data questions, first identify the missingness mechanism by looking at relationships between missingness patterns and observed variables—this determines your analytical strategy.

Question 2

A researcher studying workplace stress finds that employees with higher stress levels are more likely to have missing data for income questions on a survey. Additionally, some employees randomly failed to receive the income section due to a printing error that affected every 10th survey booklet. What type of missing data mechanism best describes the missing income data in this study?

  1. Missing Completely at Random (MCAR) because the printing error was random
  2. Missing at Random (MAR) because missingness depends on observed stress levels
  3. Missing Not at Random (MNAR) because income itself influences the missingness pattern
  4. A combination of MCAR and MNAR mechanisms operating simultaneously
  5. A combination of MCAR and MAR mechanisms operating simultaneously (correct answer)
Explanation: When you encounter questions about missing data mechanisms, you need to identify what causes the missingness and whether that cause is observed, unobserved, or completely random. This scenario presents two distinct mechanisms operating simultaneously. The printing error affects every 10th survey booklet regardless of any participant characteristics—this represents Missing Completely at Random (MCAR) because the missingness is purely due to chance and unrelated to any variables in the study. Meanwhile, employees with higher stress levels are more likely to skip income questions, which represents Missing at Random (MAR) because missingness depends on an observed variable (stress level) but not on the missing variable itself (income). Answer A is incomplete because it only acknowledges the printing error component while ignoring the stress-related missingness pattern. Answer B similarly focuses on just one mechanism, recognizing the MAR component but overlooking the MCAR printing error. Answer C mischaracterizes the situation as Missing Not at Random (MNAR), which would require income levels themselves to directly influence whether income data is missing—but the scenario indicates stress levels, not income amounts, drive the non-response pattern. Answer D correctly identifies that both MCAR and MAR mechanisms are operating together. Real-world studies frequently involve multiple missing data mechanisms simultaneously, and you must recognize all patterns present rather than forcing the situation into a single category. Remember: missing data questions often involve multiple mechanisms. Read carefully to identify all sources of missingness and classify each one separately before determining the overall pattern.

Question 3

In a longitudinal study of depression, researchers notice that participants with more severe depression scores at baseline are increasingly likely to drop out of subsequent follow-up visits. However, among participants with similar baseline depression scores, the probability of dropout appears unrelated to their unobserved future depression trajectory. What missing data mechanism does this scenario most likely represent?

  1. Missing Completely at Random (MCAR) because dropout is unrelated to future depression
  2. Missing at Random (MAR) because dropout depends on observed baseline depression scores (correct answer)
  3. Missing Not at Random (MNAR) because depression severity influences dropout decisions
  4. Cannot be determined without knowing the correlation between baseline and follow-up scores
  5. A mixture of MAR and MNAR depending on the individual participant characteristics
Explanation: Missing data mechanisms are fundamental to understanding how dropout patterns affect study validity. When analyzing longitudinal studies, you need to classify why data is missing to choose appropriate statistical methods. This scenario describes Missing at Random (MAR) because dropout depends on observed baseline depression scores, but among participants with similar baseline scores, dropout is unrelated to their future (unobserved) depression trajectory. The key insight is that once you condition on the observed baseline data, the missingness mechanism becomes random with respect to the unobserved outcomes. Answer B correctly identifies MAR. The dropout probability can be fully explained by the observed baseline depression scores, making the missing data mechanism "ignorable" for analysis purposes once baseline severity is accounted for. Answer A incorrectly suggests MCAR. While dropout is unrelated to future depression trajectories within baseline severity groups, it's not completely random since it varies by baseline depression scores. MCAR would require dropout to be independent of both observed and unobserved variables. Answer C misidentifies this as MNAR. MNAR occurs when missingness depends on the unobserved values themselves (like participants dropping out specifically because their current depression worsened). Here, dropout depends only on observed baseline scores, not unobserved future trajectories. Answer D is incorrect because the correlation between baseline and follow-up scores doesn't determine the missing data mechanism. The mechanism depends on what drives the dropout decision, not the strength of variable relationships. Remember: MAR means missingness can be explained by observed data alone. If you can predict who drops out using only baseline characteristics, you're likely dealing with MAR.

Question 4

A survey about illegal drug use has a 15% non-response rate that appears unrelated to any measured demographic variables (age, gender, education, income). However, the researchers suspect that individuals who use illegal drugs are more likely to refuse participation regardless of their demographic characteristics. If this suspicion is correct, what type of missing data mechanism would this represent?

  1. Missing Completely at Random (MCAR) because non-response appears unrelated to demographics
  2. Missing at Random (MAR) because we can predict missingness from demographic variables
  3. Missing Not at Random (MNAR) because missingness depends on the unobserved outcome of interest (correct answer)
  4. The mechanism cannot be determined from survey data alone regardless of response patterns
  5. Missing at Random (MAR) because drug use behavior is a measurable participant characteristic
Explanation: When analyzing missing data in biostatistics, you need to classify the mechanism based on whether missingness depends on observed variables, unobserved variables, or neither. This classification determines which statistical methods are appropriate and whether your results will be biased. Here, the key insight is recognizing what drives the non-response. The researchers suspect that people who actually use illegal drugs are more likely to refuse participation, regardless of their demographic characteristics. This means the probability of missing data depends directly on the unobserved value of the outcome variable itself (drug use status), making this Missing Not at Random (MNAR). Under MNAR, the missingness mechanism is related to the unobserved values, creating non-ignorable missingness that can severely bias results if not properly addressed. Option A is incorrect because MCAR requires that missingness be completely unrelated to both observed AND unobserved variables. While demographics don't predict missingness here, the unobserved drug use status does. Option B is wrong because MAR means you can predict missingness using observed variables. Since demographics don't predict non-response in this scenario, the missing data cannot be MAR. Option D is incorrect because we can determine the mechanism type based on understanding what drives the missingness, even without statistical tests. Study tip: Remember the hierarchy: MCAR (missingness unrelated to anything), MAR (missingness predictable from observed data), and MNAR (missingness depends on unobserved values). Questions about sensitive topics like drug use, income, or illegal behavior often involve MNAR mechanisms.

Question 5

In a clinical trial, blood pressure measurements are missing for some participants due to equipment malfunction on specific days, and also missing for participants who felt too unwell to attend appointments. The equipment failures occurred randomly across the study period, while the attendance-related missingness correlates with participants' measured baseline health status. What is the most appropriate way to characterize the missing data mechanism?

  1. Primarily MCAR since equipment failure is the main source of random missingness
  2. Primarily MAR since missingness can be predicted from baseline health measurements
  3. MNAR because feeling unwell represents an unmeasurable subjective state affecting missingness
  4. Mixed mechanism requiring separate analysis strategies for equipment-related versus health-related missingness (correct answer)
  5. MAR overall because both sources of missingness can be modeled using available study data
Explanation: When you encounter missing data problems in biostatistics, you need to identify the underlying mechanism because it determines which analytical approaches are valid. Missing data mechanisms fall into three categories: MCAR (missing completely at random), MAR (missing at random), and MNAR (missing not at random). This scenario presents two distinct sources of missingness with different mechanisms. The equipment failures create MCAR missingness—these occur randomly across time and participants, with no relationship to any measured or unmeasured variables. However, the attendance-related missingness operates differently. When participants miss appointments because they feel unwell, and this correlates with their baseline health status, you have MAR missingness—the missingness can be predicted from observed covariates. Option D correctly recognizes that both mechanisms operate simultaneously, requiring different analytical strategies for each type of missing data. Option A incorrectly assumes equipment failure dominates and ignores the systematic health-related missingness. Option B oversimplifies by focusing only on the MAR component while neglecting the MCAR equipment failures. Option C mischaracterizes the health-related missingness as MNAR—while feeling unwell might seem unmeasurable, the key insight is that it correlates with baseline health measurements, making it MAR rather than MNAR. The mixed mechanism scenario is common in real clinical trials. Remember to identify all sources of missingness separately rather than trying to force the entire dataset into a single category. Each mechanism may require different imputation methods or analytical approaches to maintain validity.

Question 6

A researcher tests whether data are Missing Completely at Random (MCAR) using Little's MCAR test and obtains a p-value of 0.03. The researcher concludes that the data are Missing at Random (MAR) rather than MCAR. What is the primary limitation of this reasoning?

  1. The p-value is not significant enough to reject the MCAR hypothesis definitively
  2. Little's MCAR test cannot distinguish between MAR and MNAR mechanisms reliably
  3. Rejecting MCAR does not automatically confirm MAR since the data could be MNAR (correct answer)
  4. The test requires a larger sample size to have sufficient power for detecting MCAR violations
  5. Little's MCAR test is only valid when the proportion of missing data exceeds 20%
Explanation: Missing data analysis requires understanding three distinct mechanisms: Missing Completely at Random (MCAR), Missing at Random (MAR), and Missing Not at Random (MNAR). When you encounter Little's MCAR test questions, remember that statistical tests can only tell you what to reject, not what to accept. Little's MCAR test specifically tests the null hypothesis that data are MCAR. With a p-value of 0.03 (< 0.05), you reject the MCAR hypothesis, meaning the missingness pattern is not completely random. However, rejecting MCAR leaves two possibilities: the data could be MAR (missing at random, but dependent on observed variables) or MNAR (missing not at random, dependent on unobserved values). The test cannot distinguish between these two alternatives, so concluding the data are MAR is premature and illogical. Option A is wrong because p = 0.03 is statistically significant at the conventional α = 0.05 level, providing sufficient evidence to reject MCAR. Option B mischaracterizes the test's purpose—Little's MCAR test isn't designed to distinguish MAR from MNAR; it only tests whether data deviate from MCAR. Option D incorrectly suggests a power issue when the significant result actually indicates adequate power to detect the deviation from MCAR. Study tip: Remember that statistical tests tell you what to reject, not what to conclude positively. When you reject one hypothesis in a multiple-alternative scenario, you haven't automatically confirmed any specific alternative. Always consider what remains possible after rejection.

Question 7

In a study measuring both income and job satisfaction, income data are missing more frequently for participants with low job satisfaction scores. However, when the analysis is stratified by education level, this relationship between job satisfaction and missing income data disappears within each education stratum. What does this pattern suggest about the missing data mechanism?

  1. The data are MCAR because missingness becomes random within education strata
  2. The data are MAR with education as a key covariate for modeling missingness (correct answer)
  3. The data are MNAR because income missingness relates to job satisfaction levels
  4. The relationship is confounded and the true missing data mechanism cannot be determined
  5. The data demonstrate a complex MNAR mechanism that varies by education level
Explanation: When analyzing missing data patterns, you need to understand the three key mechanisms: MCAR (Missing Completely at Random), MAR (Missing at Random), and MNAR (Missing Not at Random). The critical distinction lies in whether missingness depends on observed or unobserved variables. This scenario demonstrates a classic MAR pattern. Initially, income data appears to be missing more often for participants with low job satisfaction, suggesting the missingness isn't random. However, when you stratify by education level, this relationship disappears within each education stratum. This means education is acting as a confounding variable that explains the apparent relationship between job satisfaction and missingness. Once you account for education, the missingness becomes random within each education group. Option A is incorrect because the data aren't MCAR overall – there's still a systematic pattern of missingness that depends on observed variables (education). MCAR would mean no relationship exists at any level of analysis. Option C misinterprets the situation as MNAR, but since the relationship between job satisfaction and missingness disappears when conditioning on an observed variable (education), this isn't MNAR. MNAR would require missingness to depend on the unobserved income values themselves. Option D is wrong because we can determine the mechanism – the pattern clearly indicates MAR once we identify education as the key variable. Remember: When missingness patterns change after stratifying by observed covariates, you're likely dealing with MAR. The key is identifying which observed variables explain the missingness pattern.

Question 8

In a patient satisfaction survey, response rates vary by clinic location: Clinic A has 90% response rate, Clinic B has 75% response rate, and Clinic C has 60% response rate. Within each clinic, non-response appears unrelated to any measured patient characteristics. However, the clinics serve different patient populations with varying demographics. What missing data mechanism does this scenario represent?

  1. MCAR because non-response is random within each clinic location independently
  2. MAR because clinic location is observed and predicts response probability (correct answer)
  3. MNAR because different clinics likely have different underlying satisfaction levels
  4. Cannot be determined without knowing the demographic differences between clinic populations
  5. MCAR overall because the systematic differences are due to location, not patient characteristics
Explanation: When you encounter missing data problems in biostatistics, you need to classify the mechanism based on whether missingness depends on observed or unobserved variables. The three key mechanisms are MCAR (completely random), MAR (random conditional on observed data), and MNAR (depends on unobserved factors). In this scenario, response rates differ systematically by clinic location (90%, 75%, 60%), but within each clinic, non-response appears unrelated to patient characteristics. Since clinic location is an observed variable that predicts response probability, this represents Missing At Random (MAR). The missingness isn't completely random across the entire study, but it becomes random once you condition on (account for) the observed clinic location. Option A is incorrect because the varying response rates across clinics mean the data isn't missing completely at random overall. MCAR would require the same response rate across all clinics. Option C misses the point—while clinics may have different satisfaction levels, the missing data mechanism depends on what predicts missingness, not the outcome itself. The fact that clinic location (observed) predicts response makes this MAR, not MNAR. Option D is wrong because you have sufficient information: clinic location is observed and clearly predicts response rates, which defines MAR regardless of the specific demographic differences. Remember this pattern: if an observed variable predicts missingness but missingness appears random within levels of that variable, you're looking at MAR. The key is whether the predictor of missingness is observed (MAR) or unobserved (MNAR).

Question 9

A researcher studying academic performance finds that students with lower GPA are more likely to have missing data for self-reported study hours. The researcher argues this is MAR because GPA is observed. However, a colleague suggests it might be MNAR because students who actually study very little may be more likely to skip the question regardless of their GPA. What additional information would best help distinguish between these interpretations?

  1. The correlation coefficient between GPA and actual study hours in the complete cases
  2. Whether missingness patterns persist after controlling for other academic variables
  3. The overall proportion of missing data for the study hours question
  4. Whether students with similar GPAs but different actual study hours show different missingness rates (correct answer)
  5. The distribution of GPA scores among students with complete versus missing study hour data
Explanation: When you encounter questions about missing data mechanisms, you need to distinguish between MAR (Missing At Random) and MNAR (Missing Not At Random) by understanding what drives the missingness pattern. The key insight is that MAR means missingness depends only on observed variables, while MNAR means missingness depends on the unobserved values themselves. In this scenario, the researcher claims it's MAR because missingness relates to observed GPA. The colleague suggests it's MNAR because students who actually study little may skip the question regardless of GPA. Answer D is correct because it directly tests the MNAR hypothesis. If you compare students with similar GPAs but examine whether their actual study hours affect their likelihood of missing data, you can determine if the unobserved variable (actual study time) drives missingness. If students with identical GPAs show different missingness rates based on their true study hours, this confirms MNAR. Answer A is wrong because correlations in complete cases don't reveal what drives missingness in the incomplete cases. Answer B is incorrect because controlling for other academic variables still wouldn't distinguish whether actual study hours (the missing variable itself) affects missingness. Answer C is wrong because the overall proportion of missing data tells you nothing about the mechanism causing the missingness. Remember this pattern: To distinguish MAR from MNAR, look for tests that examine whether the unobserved values themselves influence the probability of being missing, holding observed covariates constant.

Question 10

In an online survey about internet usage habits, the survey software randomly crashes for 8% of participants during completion, and separately, participants who spend more time online are more likely to abandon the survey before reaching questions about problematic internet use. What is the most accurate characterization of the missing data mechanism for the problematic internet use questions?

  1. MCAR because software crashes are random events unrelated to participant characteristics
  2. MAR because online time is measurable and predicts survey completion patterns
  3. MNAR because participants with problematic internet use avoid relevant questions
  4. Combined MCAR and MNAR mechanisms affecting the same outcome variable (correct answer)
  5. Combined MCAR and MAR mechanisms affecting the same outcome variable
Explanation: When analyzing missing data mechanisms in surveys, you need to consider whether the probability of missingness depends on observed data, unobserved data, or neither. This question presents a scenario with two distinct causes of missing data operating simultaneously. The correct answer is D because you have two different mechanisms creating missingness for the same outcome variable. The software crashes represent a Missing Completely At Random (MCAR) mechanism since technical failures occur independently of participant characteristics or responses. Meanwhile, participants with problematic internet use avoiding relevant questions represents a Missing Not At Random (MNAR) mechanism because the missingness depends on the unobserved values themselves—those with problematic use are systematically more likely to have missing data on that topic. Option A is incorrect because while software crashes are indeed MCAR, this ignores the systematic avoidance by heavy internet users. Option B mischaracterizes the avoidance mechanism as Missing At Random (MAR)—even though online time might be measurable, the avoidance specifically relates to the unobserved problematic use patterns, not just general online time. Option C incorrectly dismisses the random software crashes, focusing only on the systematic avoidance. When you encounter missing data questions, always ask: "What causes the missingness?" If multiple distinct causes operate on the same variable—especially when they follow different mechanisms—the overall pattern reflects a combination of those mechanisms. Look for scenarios where technical issues (usually MCAR) combine with systematic participant behavior (often MNAR).

Question 11

A study examining the relationship between stress and sleep quality finds that participants with missing sleep quality data have significantly higher measured stress levels than those with complete data. Additionally, participants with missing stress data have significantly lower reported sleep quality scores than those with complete stress data. What does this pattern of cross-variable relationships suggest about the missing data mechanism?

  1. MCAR because both variables show systematic patterns that cancel each other out
  2. MAR because missingness in each variable can be predicted from the other observed variable (correct answer)
  3. MNAR because participants are selectively avoiding questions about their primary problem area
  4. The pattern is inconsistent and suggests measurement error rather than systematic missingness
  5. MAR for sleep quality and MNAR for stress measurements based on different mechanisms
Explanation: When you encounter questions about missing data mechanisms, focus on the relationship between missingness patterns and observed variables. The three key mechanisms are MCAR (missing completely at random), MAR (missing at random), and MNAR (missing not at random). In this scenario, there's a clear cross-variable pattern: people missing sleep data have higher stress levels, and people missing stress data have lower sleep quality scores. This suggests that missingness in one variable can be predicted from the observed values of the other variable. When you can predict missingness based on other observed data in your dataset, this defines MAR (missing at random). The term "at random" is somewhat misleading here – it means random conditional on the observed data, not completely random. Option A is incorrect because MCAR would show no systematic relationship between missingness and any observed variables – the patterns wouldn't "cancel out" but simply wouldn't exist. Option C (MNAR) would occur if missingness depended on the unobserved values themselves (like people with severe insomnia refusing to answer sleep questions), but here the missingness is explained by the other observed variable. Option D misses the point entirely – this is a clear, consistent pattern that indicates a specific missing data mechanism, not measurement error. Remember this key distinction: MAR means you can predict missingness from other observed variables in your dataset, while MCAR means no such predictive relationships exist. Look for cross-variable patterns to identify MAR situations.

Question 12

In a longitudinal study, researchers observe that participants who miss the 6-month follow-up visit are more likely to also miss the 12-month visit, even after controlling for all baseline characteristics. Among participants who attended the 6-month visit, attendance at 12-month follow-up appears random. What missing data mechanism is most likely operating for the 12-month visit data?

  1. MCAR because 12-month attendance is random among 6-month attendees
  2. MAR because 6-month attendance status is observed and predicts 12-month attendance (correct answer)
  3. MNAR because there appears to be an unmeasured factor influencing repeated non-attendance
  4. The mechanism cannot be determined without knowing why participants missed visits
  5. MAR initially becoming MCAR after conditioning on previous attendance patterns
Explanation: When you encounter missing data questions in biostatistics, focus on what information predicts missingness and whether that information is observed or unobserved. Missing data mechanisms depend on the relationship between missingness and other variables. Here, the key insight is that 6-month attendance status is observed and strongly predicts 12-month attendance, even after controlling for baseline characteristics. This creates a situation where missingness depends on observed data. Answer B is correct because this scenario perfectly describes Missing at Random (MAR). Under MAR, the probability of missing data can depend on observed variables—in this case, whether someone attended the 6-month visit. Since we know each participant's 6-month attendance status, we have the information needed to account for this pattern of missingness in our analysis. Answer A misinterprets the situation by focusing only on the subset who attended at 6 months, ignoring the overall pattern. Answer C suggests Missing Not at Random (MNAR), but MNAR requires that missingness depends on unobserved factors. Here, the predictor (6-month attendance) is fully observed. Answer D incorrectly assumes we need to know participants' reasons for missing visits, but missing data mechanisms are defined by statistical relationships, not underlying motivations. Study tip: For missing data questions, always ask: "What predicts missingness, and is that predictor observed?" If the predictor is observed, it's MAR. If it's unobserved or the missing values themselves, it's MNAR. True randomness (MCAR) is rare in real studies.

Question 13

A researcher collects data on both self-reported and objectively measured physical activity. Self-reported data are missing more frequently among participants with low objective measurements, but objective data are missing completely at random due to device failures. When analyzing the relationship between self-reported and objective measures, what is the most important consideration regarding missing data?

  1. Use only complete cases since MCAR data support unbiased complete-case analysis
  2. The MAR mechanism for self-reported data allows for valid imputation using objective measures (correct answer)
  3. The mixed mechanisms require separate imputation strategies for each variable type
  4. MNAR patterns in self-reported data make any imputation approach potentially biased
  5. The MCAR objective data can be used to validate imputation models for self-reported data
Explanation: When analyzing missing data patterns, you need to identify the underlying mechanisms and choose strategies that leverage available information while minimizing bias. In this scenario, you have two distinct missing data mechanisms. The objective measurements are missing completely at random (MCAR) due to random device failures - the missingness has no relationship to any observed or unobserved variables. The self-reported data follows a missing at random (MAR) pattern - it's systematically missing based on an observed variable (low objective activity levels), but not based on the unobserved self-reported values themselves. The key insight is that MAR data can be validly imputed using observed variables that predict the missingness pattern. Since participants with low objective measurements are more likely to have missing self-reported data, you can use the objective measurements as predictors in your imputation model. This leverages the systematic relationship to produce unbiased estimates. Option A is incorrect because having some MCAR data doesn't justify complete-case analysis when you also have MAR data - you'd lose valuable information and introduce bias. Option C misses the point that mixed mechanisms can often be handled with a unified approach that accounts for both patterns. Option D is wrong because this isn't missing not at random (MNAR) - the missingness depends on observed objective measures, not unobserved self-reported values. Remember: MAR data is your friend in imputation because you can model the missingness using observed predictors. Always look for variables that explain why data might be missing.

Question 14

In a study of medication adherence, patients with poor adherence are more likely to skip follow-up appointments where adherence would be measured. However, the researchers have access to pharmacy refill records that accurately reflect adherence patterns for all patients, including those who miss appointments. How should the missing clinic-based adherence data be classified?

  1. MNAR because poor adherence leads to avoidance of adherence-related measurements
  2. MAR because pharmacy records provide observed data that predict missingness patterns
  3. MCAR because the relationship between adherence and missingness is now observable
  4. The classification depends on whether clinic and pharmacy measures are highly correlated
  5. MNAR initially, but effectively MAR given the availability of pharmacy data (correct answer)
Explanation: When encountering missing data problems in biostatistics, you need to classify the missingness mechanism to choose appropriate analytical methods. The key is understanding what information determines whether data is missing. Missing data mechanisms depend on whether the probability of missingness can be explained by observed data. In this scenario, patients with poor adherence avoid clinic visits where adherence would be measured, but pharmacy records capture adherence patterns for all patients. Since you have complete pharmacy data that accurately reflects the same underlying variable (adherence), you can fully observe the factor that drives missingness patterns. The correct answer is B - this is Missing at Random (MAR) because the pharmacy records provide observed data that can predict and explain the missingness patterns in clinic-based measurements. When you have observed variables that explain why data is missing, the mechanism is MAR. Answer A incorrectly suggests MNAR (Missing Not at Random). MNAR occurs when missingness depends on unobserved values of the missing variable itself. Here, the adherence patterns aren't unobserved - they're captured in pharmacy records. Answer C misapplies MCAR (Missing Completely at Random). MCAR means missingness is unrelated to any variables, observed or unobserved. The relationship between adherence and missingness clearly isn't random. Answer D unnecessarily focuses on correlation between measures. The classification depends on whether you can observe the factors driving missingness, not on how well different measures correlate. Study tip: For missing data classification, ask "Can I observe what's causing the missingness?" If yes, it's MAR. If the missing values themselves drive missingness, it's MNAR.

Question 15

A researcher studying social media use finds that participants who report higher daily usage are more likely to have missing data for questions about negative impacts of social media use. However, when the analysis includes objective phone usage data (screen time), the relationship between self-reported usage and missingness becomes non-significant, while objective usage strongly predicts missingness. What does this suggest about the missing data mechanism?

  1. MNAR because high usage leads to avoidance of questions about negative impacts
  2. MAR with objective usage as the key predictor rather than self-reported usage
  3. MCAR after controlling for objective usage measurements in the analysis
  4. The mechanism changes from MNAR to MAR when objective data are included
  5. MAR throughout, with objective usage being a more accurate predictor than self-report (correct answer)
Explanation: When analyzing missing data mechanisms, you need to understand how the probability of missingness relates to observed and unobserved variables. This question tests whether you can distinguish between different missing data mechanisms when additional information becomes available. The scenario reveals a classic case of confounding in missing data analysis. Initially, self-reported usage appears to predict missingness, suggesting participants with high usage avoid answering sensitive questions (MNAR). However, when objective phone data is included, self-reported usage becomes non-significant while objective usage strongly predicts missingness. This indicates that the true mechanism is MAR - missingness depends on an observed variable (objective usage) rather than the unobserved responses themselves. Answer A incorrectly assumes the initial relationship represents the true mechanism, ignoring how the relationship changes when better predictors are included. Answer C is wrong because the data was never MCAR - there's a clear systematic pattern of missingness related to usage levels. Answer D misunderstands that we're not observing a mechanism change, but rather uncovering the true mechanism that was previously obscured by measurement issues. The correct answer is B. The missing data mechanism is MAR, with objective usage serving as the key predictor rather than potentially biased self-reported measures. The objective data reveals that missingness depends on actual usage behavior, which is observable, making this a MAR situation. Remember: when evaluating missing data mechanisms, always consider whether apparent MNAR patterns might actually be MAR once you account for all relevant observed variables. Better measurement often clarifies the true underlying mechanism.

Question 16

In a multi-site clinical trial, Site A has 10% missing outcome data, Site B has 15% missing data, and Site C has 30% missing data. Investigation reveals that Site C had staffing shortages during data collection, while Sites A and B had adequate staffing. Within each site, missingness appears unrelated to patient characteristics or treatment assignment. What is the most appropriate missing data classification for the primary analysis?

  1. MCAR because missingness is random within each site and unrelated to outcomes
  2. MAR because site is observed and predicts missingness probability systematically (correct answer)
  3. MNAR because operational factors like staffing may correlate with unmeasured site characteristics
  4. Mixed mechanism requiring site-specific analysis approaches for valid inference
  5. MAR for Sites A and B, but MNAR for Site C due to operational differences
Explanation: When analyzing missing data patterns in biostatistics, you need to classify the mechanism to choose appropriate analytical methods. The three main categories are MCAR (Missing Completely at Random), MAR (Missing at Random), and MNAR (Missing Not at Random). The correct answer is B because site is an observed variable that systematically predicts missingness probability. Site C's 30% missing rate due to staffing shortages creates a clear pattern where knowing the site tells you the likelihood of missing data. Since site is observed and can be included in your analysis model, this makes the missingness MAR. The key insight is that while missingness appears random within each site, it's systematically different between sites due to the operational factor (staffing). Option A is incorrect because MCAR requires missingness to be completely random across the entire study, not just within sites. The dramatic difference between sites (10%, 15%, vs 30%) violates this assumption. Option C misapplies MNAR, which occurs when missingness depends on the unobserved outcome itself. Here, missingness depends on an observed variable (site) and operational factors, not on unmeasured patient outcomes. Option D overcomplicates the situation. While different sites have different missing rates, this doesn't necessarily require different analytical approaches since site can be incorporated as a covariate in standard MAR methods. Remember: MAR doesn't mean "random everywhere" – it means missingness is random conditional on observed variables. If an observed factor (like study site) predicts missingness patterns, classify it as MAR and include that factor in your analysis.

Question 17

A survey about workplace harassment includes a question about whether the respondent has personally experienced harassment. This question has a 25% non-response rate, while other sensitive questions (salary, personal relationships) have only 5% non-response rates. What does this pattern most strongly suggest about the missing data mechanism for the harassment question?

  1. MCAR because all sensitive questions have some level of non-response
  2. MAR because non-response can be predicted from sensitivity levels of different questions
  3. MNAR because individuals who experienced harassment may be more likely to skip this specific question (correct answer)
  4. The mechanism cannot be determined without demographic predictors of non-response
  5. MAR because other sensitive questions provide a baseline for expected non-response rates
Explanation: When you encounter questions about missing data patterns, focus on what drives the non-response behavior. Missing data mechanisms fall into three categories: MCAR (missing completely at random), MAR (missing at random), and MNAR (missing not at random). The key is understanding whether the "missingness" itself relates to the unobserved values. The dramatically higher non-response rate for the harassment question (25% vs. 5%) suggests that people's decision to skip this question is directly related to their harassment experience. Those who experienced harassment are likely more hesitant to answer due to trauma, fear of identification, or emotional difficulty, while those who haven't experienced harassment may feel comfortable answering. This creates a systematic relationship between the missing data and the unobserved values themselves—the hallmark of MNAR. Option A is wrong because MCAR requires missingness to be completely random, unrelated to any variable. The pattern here is clearly systematic, not random. Option B incorrectly suggests MAR, which would mean non-response could be predicted from other observed variables (like demographics), but the evidence points to the harassment experience itself driving non-response. Option D misses the point—while demographic predictors might help confirm the mechanism, the stark difference in response rates already provides strong evidence of the underlying mechanism. Remember: MNAR occurs when the probability of missing data depends on the unobserved values themselves. In surveys about sensitive personal experiences, be especially alert for MNAR patterns where those most affected by the topic are most likely to avoid responding.

Question 18

A longitudinal study tracks both self-reported pain levels and objective biomarkers of inflammation. At each time point, participants with missing self-reported pain data show significantly higher inflammation biomarkers than those with complete pain data. However, participants with missing biomarker data show inflammation levels that are indistinguishable from the overall sample when measured at adjacent time points. How should the missing data mechanisms be characterized?

  1. MAR for both variables because each can be predicted from available measurements
  2. MAR for pain data and MCAR for biomarker data based on different predictability patterns (correct answer)
  3. MNAR for pain data because high inflammation suggests severe pain that leads to non-reporting
  4. MCAR for both variables after accounting for the longitudinal correlation structure
  5. Mixed longitudinal mechanism requiring time-varying missing data models
Explanation: When analyzing missing data in longitudinal studies, you need to distinguish between three mechanisms: MCAR (Missing Completely At Random), MAR (Missing At Random), and MNAR (Missing Not At Random). The key is examining whether missingness patterns relate to observed or unobserved variables. For the pain data, participants with missing responses show significantly higher inflammation biomarkers. This creates a clear pattern where missingness is predictable from observed inflammation levels, indicating MAR. The missing pain reports aren't random—they're systematically related to high inflammation, which you can observe and account for in your analysis. For the biomarker data, participants with missing values show inflammation levels indistinguishable from the overall sample at adjacent time points. This suggests no systematic pattern—the missingness appears unrelated to any observable characteristics, indicating MCAR. Option A incorrectly assumes both variables follow the same mechanism simply because each could theoretically be predicted. Option C suggests MNAR for pain data, but MNAR would require the missingness to depend on the unobserved pain values themselves, not the observed inflammation markers. Option D incorrectly assumes longitudinal correlation automatically makes missing data MCAR across both variables. The correct characterization is B: MAR for pain data (predictable from observed inflammation) and MCAR for biomarker data (no observable pattern). Study tip: Always examine what observed variables relate to missingness patterns. If you can predict missingness from observed data, it's MAR. If there's no detectable pattern, consider MCAR. Reserve MNAR for when missingness depends on the unobserved values themselves.

Question 19

A nutritionist studying eating habits finds that participants with eating disorders are more likely to have missing data for questions about caloric intake, but not for questions about exercise habits. Among participants without eating disorders, missing data patterns appear random for both types of questions. How should this missing data pattern be characterized?

  1. MCAR for exercise data and MAR for caloric intake data across all participants
  2. MNAR for caloric intake data because eating disorder status affects missingness patterns
  3. MAR overall because eating disorder status is observed and can predict missingness (correct answer)
  4. Mixed mechanism with MCAR within subgroups but MAR when groups are combined
  5. MNAR for both variables because eating disorders influence all aspects of dietary reporting
Explanation: When analyzing missing data patterns, you need to consider three key mechanisms: MCAR (Missing Completely At Random), MAR (Missing At Random), and MNAR (Missing Not At Random). The critical distinction lies in whether you can predict missingness from observed variables. In this scenario, eating disorder status is an observed variable that predicts missing caloric intake data. Since you can identify which participants are more likely to have missing data based on this observable characteristic, the missingness follows a predictable pattern that can be accounted for in your analysis. This defines MAR - the missingness depends on observed covariates but not on the unobserved values themselves. Answer A incorrectly splits the mechanisms by question type rather than recognizing that the overall pattern across all participants is MAR. While exercise data might appear MCAR within each group, the presence of eating disorder status as a predictor makes the overall mechanism MAR. Answer B suggests MNAR, which would apply if the missing caloric intake data depended on the actual caloric values themselves (e.g., people with extremely high intake refusing to report). However, since eating disorder status is observed and explains the missingness, this is MAR, not MNAR. Answer D overcomplicates the classification. While the description might seem accurate, the standard approach is to characterize the overall mechanism as MAR when an observed variable predicts missingness patterns. Remember: if you can predict who has missing data using variables you've observed, it's MAR. This distinction is crucial because MAR allows for valid statistical correction methods.

Question 20

In a study examining both anxiety and depression symptoms, researchers find that participants with missing anxiety data tend to have higher observed depression scores, while participants with missing depression data tend to have higher observed anxiety scores. Both patterns persist after controlling for demographic variables. What missing data mechanism does this cross-symptom pattern most likely represent?

  1. MCAR because the cross-relationships create a balanced pattern of missingness
  2. MAR because missingness in each variable can be predicted from the other observed variable (correct answer)
  3. MNAR because participants avoid reporting their more severe symptom domain
  4. Mixed mechanisms with MAR for anxiety and MNAR for depression based on different reporting patterns
  5. Cannot be determined because the cross-relationships create ambiguous mechanism classification
Explanation: When analyzing missing data patterns in biostatistics, you need to understand the three fundamental mechanisms: MCAR (Missing Completely at Random), MAR (Missing at Random), and MNAR (Missing Not at Random). The key distinction lies in whether missingness depends on observed data, unobserved data, or neither. In this scenario, the cross-symptom pattern reveals a classic MAR mechanism. People missing anxiety data have higher observed depression scores, and people missing depression data have higher observed anxiety scores. This means the missingness in each variable is systematically related to the observed values of the other variable. Since you can predict who will have missing anxiety data based on their observed depression scores (and vice versa), this missingness is "random" conditional on the observed information—the definition of MAR. Option A is incorrect because MCAR would show no systematic relationships between missingness and any observed variables. The clear patterns described rule this out. Option C misinterprets the mechanism—if people were avoiding reporting their more severe symptoms, you'd expect those with missing data to have higher scores in that same domain when observed, not the opposite pattern described. Option D incorrectly assumes different mechanisms are operating when the symmetric cross-pattern suggests a single MAR mechanism affecting both variables. Study tip: For missing data questions, always ask: "Can I predict who has missing data based on what I can observe?" If yes, it's likely MAR. The cross-variable prediction pattern here is a telltale MAR signature.