Summarize this pdf for rmy research examination and give me 30 research based mcqs
Reading File
Running Command
python3 << 'EOF' import pdfplumber with pdfplumber.open("/tmp/workspace/attachments/c9de8ed7-9f5b-4044-90f2-15ddad65bd6f/RS1 Errors and Bias.pdf") as pdf: print(f"Total pages: {len(pdf.pages)}") full_text = "" for i, page in enumerate(pdf.pages): text = page.extract_text() if text: full_text += f"\n--- PAGE {i+1} ---\n{text}" print(full_text) EOF
I've extracted all 18 pages of the PDF. Now I'll write a full summary and 30 MCQs based on the content.
RS1: Research Errors & Bias - Summary
1. What is Error?
Error refers to the deviation of observed results from the true value due to imperfections in measurement, sampling, or analysis. Error can be either systematic or random, and both affect:
- Accuracy (validity) - how close results are to the true value
- Precision (reliability) - how consistent results are
2. Five Types of Research Error
| Type | Description | Example |
|---|---|---|
| Random Error | Unpredictable fluctuations that scatter results in both directions; reduces precision but not systematically wrong | Variation in blood pressure readings |
| Systematic Error | Results consistently deviate in one direction due to flaws in instruments or methods - essentially bias | A miscalibrated scale always reading 2 kg heavy |
| Measurement Error | Inaccuracy in recording variables; can be random (noisy) or systematic (consistent misreporting) | Participants underreporting alcohol consumption |
| Sampling Error | Arises because studies use samples rather than entire populations; the difference between sample estimate and true population value is due to chance | A small sample by chance over-representing elderly patients |
| Human Error | Mistakes in recording, entering, or interpreting data due to human oversight; may be systematic or random | Researcher miskeys "150" instead of "105" for blood sugar |
3. Selection Bias
Occurs when individuals chosen for a study are not representative of the target population, distorting associations between exposure and outcome.
| Subtype | Description |
|---|---|
| Sampling Bias | Flawed selection produces unrepresentative groups |
| Exclusion Bias | Systematic omission of specific groups |
| Survivor Bias | Only survivors included, ignoring poor outcomes |
| Healthy Worker Effect | Employed individuals are healthier than the general population |
| Volunteer Bias | Participants who choose to join differ systematically - often more motivated or health-conscious |
| Ascertainment Bias | Certain groups more likely to be detected due to surveillance or healthcare access differences |
| Attrition Bias | Dropouts differ meaningfully from those who remain, skewing longitudinal results (e.g., sicker diabetics drop out, making treatment appear more effective) |
| Referral Bias | Patients at specialized centers differ from the broader population, often with more severe disease |
4. Information (Measurement) Bias
Systematic errors in how data about exposures, outcomes, or variables are collected, recorded, or classified. These consistently skew data in one direction.
| Subtype | Description |
|---|---|
| Recall Bias | Patients with disease remember past exposures differently than healthy controls |
| Interviewer Bias | Data collector influences responses through wording, tone, or assumptions |
| Observer Bias | Researcher's expectations influence how they record observations |
| Response Bias | Participants provide inaccurate answers due to social desirability or embarrassment on sensitive topics |
| Instrument Bias | Measurement tool is systematically inaccurate or poorly calibrated |
| Detection Bias | Outcomes identified more frequently in one group due to differences in monitoring intensity |
| Misclassification Bias | Participants placed into wrong exposure or outcome categories - can be differential or non-differential, diluting or exaggerating associations |
5. Publication and Reporting Biases
| Subtype | Description |
|---|---|
| Publication Bias | Studies with positive results are more likely published than those with null findings, creating a distorted evidence base |
| Time-Lag Bias | Positive findings publish faster than negative ones, giving early readers an inflated impression of benefit |
| Citation Bias | Positive studies cited more frequently while negative studies are ignored, reinforcing one-sided evidence |
| Outcome Reporting Bias | Only favorable or significant outcomes are reported while others measured in the study are suppressed |
| Language Bias | Studies selectively published in certain languages based on results direction (positive findings in English journals, null results in regional publications) |
6. Confounding
Occurs when a third variable related to both exposure and outcome distorts the observed association, giving a misleading impression of causality.
- Classic Example: Coffee drinking and lung cancer - smoking is the confounder because smokers drink more coffee AND develop more lung cancer. Without adjustment, coffee falsely appears to cause cancer.
| Subtype | Description |
|---|---|
| Confounding Bias | Primary exposure effect is mixed with another variable related to both exposure and outcome but not part of the causal pathway |
| Residual Confounding | Confounding persists after control attempts because the confounder was measured inaccurately, incompletely, or not at all |
| Confounding by Indication | The reason for prescribing treatment is itself related to the outcome, making it difficult to separate treatment effect from underlying condition |
| Overadjustment Bias | Adjusting for intermediates on the causal pathway removes part of the real effect, underestimating the true relationship |
7. Performance and Attrition Bias (in Experimental Trials)
- Performance Bias: Participants in different groups receive unequal care, attention, or co-interventions beyond the study treatment, distorting the true intervention effect.
- Detection Bias: Outcome assessment influenced by knowledge of participant group assignment - especially problematic for subjective outcomes like pain or quality of life.
- Blinding of participants, staff, and outcome assessors is the primary defense against these biases in RCTs.
Attrition Bias Types:
- Attrition Bias - Systematic differences between groups in withdrawal, dropout, or exclusion
- Differential Attrition - Dropout rate differs between experimental and control groups, undermining randomization
- Improper Handling - Analyzing only completers ("per-protocol") instead of all randomized participants ("intention-to-treat") inflates benefits by excluding poor responders
8. Behavioral and Expectancy Biases
| Subtype | Description |
|---|---|
| Hawthorne Effect | Participants change behavior because they know they're being observed, not due to the intervention itself |
| Confirmation Bias | Researchers interpret data to confirm pre-existing beliefs, affecting analysis and which findings are emphasized |
| Observer-Expectancy | Researcher expectations subtly influence participant behavior or data recording (e.g., "maze-bright" rats recorded better performance due to expectations) |
9. Analysis and Design Biases
| Subtype | Description |
|---|---|
| Channeling Bias | Certain patients systematically steered toward particular interventions based on prognostic factors, creating baseline differences between treatment groups |
| Surveillance Bias | One group monitored more closely than another, leading to higher diagnosis rates due to increased observation, not true differences |
| Data Dredging | Repeatedly testing many hypotheses until finding significant results, then selectively reporting them - inflates false-positive findings |
| Analysis Bias | Inappropriate statistical methods or selective analytic choices distort results, including not adjusting for clustering or misusing models |
30 Research-Based MCQs
Q1. A researcher measures blood pressure in 50 patients using a poorly calibrated sphygmomanometer that consistently reads 10 mmHg higher than the true value. This best represents:
- A) Random error
- B) Sampling error
- C) Systematic error
- D) Human error
Answer: C - The error is consistent and in one direction, which is the defining feature of systematic error.
Q2. A study recruits participants via social media and finds that volunteers are more health-conscious than the general population. This is an example of:
- A) Referral bias
- B) Volunteer bias
- C) Ascertainment bias
- D) Attrition bias
Answer: B - Volunteer bias occurs when those who choose to participate differ systematically from those who decline.
Q3. In a case-control study on lung cancer, cases recall past smoking habits more accurately than controls. This differential recall introduces:
- A) Observer bias
- B) Recall bias
- C) Interviewer bias
- D) Detection bias
Answer: B - Recall bias occurs when diseased individuals remember past exposures differently than healthy controls.
Q4. A study on occupational health finds that employed workers have lower mortality than the general population, even in hazardous industries. This is best explained by:
- A) Survivor bias
- B) Channeling bias
- C) Healthy worker effect
- D) Referral bias
Answer: C - The healthy worker effect describes how employed individuals are healthier than the general population due to selection at hiring.
Q5. A randomized controlled trial finds that the treatment group receives more attention from nursing staff due to perceived intervention importance. The observed benefit may be inflated by:
- A) Attrition bias
- B) Detection bias
- C) Performance bias
- D) Hawthorne effect
Answer: C - Performance bias occurs when groups receive unequal co-interventions or care beyond the study treatment.
Q6. Which of the following best describes the Hawthorne effect?
- A) Researchers confirm pre-existing beliefs when interpreting data
- B) Participants alter behavior because they know they are being observed
- C) Dropout rates differ between study arms
- D) Outcomes are detected more often in a closely monitored group
Answer: B - The Hawthorne effect is when awareness of observation (not the intervention) changes participant behavior.
Q7. A pharmaceutical trial publishes only the outcome showing significant weight reduction and omits non-significant blood pressure and cholesterol changes. This is:
- A) Publication bias
- B) Language bias
- C) Outcome reporting bias
- D) Citation bias
Answer: C - Outcome reporting bias occurs when only favorable or significant outcomes from a study are reported, while others are suppressed.
Q8. Coffee drinking is found to be associated with lung cancer. Upon further analysis, smoking explains the entire association. Smoking is best described as:
- A) An effect modifier
- B) A confounder
- C) A mediator
- D) A collider
Answer: B - Smoking is related to both the exposure (coffee drinking) and the outcome (lung cancer) and is not on the causal pathway - the definition of a confounder.
Q9. A researcher repeatedly tests multiple subgroups in a dataset until finding a p-value < 0.05, then reports only that result. This is known as:
- A) Analysis bias
- B) Channeling bias
- C) Data dredging
- D) Outcome reporting bias
Answer: C - Data dredging (p-hacking) involves testing many hypotheses until a significant result is found, then selectively reporting it.
Q10. In a longitudinal diabetes trial, sicker patients drop out at a higher rate from the treatment arm. The resulting appearance of improved treatment efficacy demonstrates:
- A) Referral bias
- B) Attrition bias
- C) Surveillance bias
- D) Volunteer bias
Answer: B - Attrition bias occurs when dropout differs meaningfully between groups; here sicker patients leaving inflates apparent treatment benefit.
Q11. Analyzing only patients who completed a trial (per-protocol analysis) rather than all who were randomized (intention-to-treat) is most problematic because it:
- A) Increases random error
- B) Introduces sampling error
- C) Excludes poor responders and inflates benefit
- D) Reduces statistical power
Answer: C - Per-protocol analysis excludes dropouts who often respond poorly, thus overestimating treatment benefit compared to intention-to-treat analysis.
Q12. An interviewer studying domestic violence unconsciously uses a softer tone when speaking with participants suspected of abuse. This represents:
- A) Observer bias
- B) Confirmation bias
- C) Interviewer bias
- D) Response bias
Answer: C - Interviewer bias occurs when the data collector's wording, tone, or assumptions influence participant responses.
Q13. A study on herbal therapy shows positive results in English-language journals but null results appear only in regional non-English publications. This is an example of:
- A) Outcome reporting bias
- B) Language bias
- C) Time-lag bias
- D) Citation bias
Answer: B - Language bias is when studies are selectively published in certain languages based on the direction of results.
Q14. Positive findings in a meta-analysis are published within 6 months while negative findings take 3 years to appear. This describes:
- A) Publication bias
- B) Time-lag bias
- C) Citation bias
- D) Language bias
Answer: B - Time-lag bias is the faster publication of positive findings, giving early readers an inflated impression of benefit.
Q15. A new antihypertensive is preferentially prescribed to younger, lower-risk patients, making it appear safer in observational data. This is:
- A) Confounding by indication
- B) Channeling bias
- C) Overadjustment bias
- D) Residual confounding
Answer: B - Channeling bias occurs when patients are systematically steered toward particular interventions based on prognostic factors, creating baseline differences.
Q16. A researcher adjusting for BMI in a study of diet and cardiovascular disease inadvertently removes part of the true biological effect since BMI mediates the relationship. This is:
- A) Residual confounding
- B) Confounding by indication
- C) Overadjustment bias
- D) Channeling bias
Answer: C - Overadjustment bias occurs when adjustment for intermediates on the causal pathway underestimates the true effect.
Q17. After adjusting for all measured confounders, a residual association persists because the confounder was measured imprecisely. This is:
- A) Confounding bias
- B) Residual confounding
- C) Overadjustment bias
- D) Channeling bias
Answer: B - Residual confounding persists after control attempts because the confounder was measured inaccurately, incompletely, or not at all.
Q18. A scale used in a study is poorly calibrated and consistently underestimates weight by 3 kg for all participants. This is:
- A) Random measurement error
- B) Sampling error
- C) Instrument bias
- D) Human error
Answer: C - Instrument bias occurs when a measurement tool is systematically inaccurate or poorly calibrated, causing consistent errors.
Q19. Participants in a study on alcohol intake report fewer drinks per week than they actually consume. This systematic underreporting is an example of:
- A) Recall bias
- B) Response bias
- C) Misclassification bias
- D) Observer bias
Answer: B - Response bias occurs when participants provide inaccurate answers due to social desirability or embarrassment on sensitive topics.
Q20. Patients referred to a tertiary cardiac center tend to have more severe disease than the general population with the same diagnosis. Using only this center's data introduces:
- A) Volunteer bias
- B) Referral bias
- C) Exclusion bias
- D) Ascertainment bias
Answer: B - Referral bias occurs because patients at specialized centers differ from the broader population, often with more severe disease.
Q21. Which of the following is the primary defense against performance bias and detection bias in RCTs?
- A) Stratified randomization
- B) Intention-to-treat analysis
- C) Blinding of participants, staff, and outcome assessors
- D) Large sample size
Answer: C - Blinding prevents both performance bias (unequal care) and detection bias (knowledge of group assignment affecting outcome assessment).
Q22. A cancer registry detects more cases in a high-income group because they access healthcare more frequently, not because they have higher true incidence. This is:
- A) Surveillance bias
- B) Ascertainment bias
- C) Detection bias
- D) Healthy worker effect
Answer: B - Ascertainment bias occurs when certain groups are more likely to be detected due to surveillance or healthcare access differences.
Q23. In a study on pain management, an unblinded assessor consistently scores pain lower in the treatment group because they expect better outcomes. This is:
- A) Observer-expectancy bias
- B) Hawthorne effect
- C) Confirmation bias
- D) Performance bias
Answer: A - Observer-expectancy bias occurs when researcher expectations subtly influence their recording of data or participant behavior.
Q24. Researchers interpret ambiguous data in a way that supports their original hypothesis about a new drug's effectiveness. This is:
- A) Observer-expectancy bias
- B) Confirmation bias
- C) Hawthorne effect
- D) Channeling bias
Answer: B - Confirmation bias occurs when researchers interpret data to confirm pre-existing beliefs, affecting analysis and which findings are emphasized.
Q25. Which type of error is defined as unpredictable fluctuations that scatter results in both directions, reducing precision but not causing systematic distortion?
- A) Systematic error
- B) Random error
- C) Measurement error
- D) Sampling error
Answer: B - Random error produces scatter in both directions (no consistent direction), reducing precision but not validity.
Q26. A study excludes patients with comorbidities, leaving out a group that has the worst outcomes. This omission is:
- A) Volunteer bias
- B) Survivor bias
- C) Exclusion bias
- D) Attrition bias
Answer: C - Exclusion bias is the systematic omission of specific groups from study participation.
Q27. In a study comparing two surgical techniques, the group receiving the newer surgery has significantly more dropouts due to side effects. This threatens internal validity through:
- A) Differential attrition
- B) Survivor bias
- C) Instrument bias
- D) Surveillance bias
Answer: A - Differential attrition occurs when the dropout rate differs between the experimental and control groups, undermining the benefits of randomization.
Q28. Participants assigned to an exercise intervention walk more than usual simply because they know they are part of a trial. This most likely reflects:
- A) Observer-expectancy bias
- B) Hawthorne effect
- C) Performance bias
- D) Volunteer bias
Answer: B - The Hawthorne effect describes participants changing behavior because they know they are being observed or studied.
Q29. A study uses inappropriate statistical models that ignore clustering of patients within hospitals, distorting the estimated effect of a hospital-level intervention. This is:
- A) Data dredging
- B) Channeling bias
- C) Analysis bias
- D) Surveillance bias
Answer: C - Analysis bias results from inappropriate statistical methods or selective analytic choices that distort results, including not adjusting for clustering.
Q30. Which of the following scenarios best illustrates misclassification bias?
- A) A researcher reads the outcome knowing which arm the patient was in
- B) An exposed participant is incorrectly categorized as unexposed due to a data entry error
- C) A hospital only enrolls patients with severe disease
- D) Subjects report less alcohol intake than they truly consume
Answer: B - Misclassification bias occurs when participants are placed in the wrong exposure or outcome categories, diluting or exaggerating associations.
Quick Reference: Bias Categories
| Category | Key Biases |
|---|---|
| Selection | Sampling, Exclusion, Survivor, Healthy Worker, Volunteer, Ascertainment, Attrition, Referral |
| Information/Measurement | Recall, Interviewer, Observer, Response, Instrument, Detection, Misclassification |
| Publication/Reporting | Publication, Time-lag, Citation, Outcome Reporting, Language |
| Confounding | Confounding Bias, Residual Confounding, Confounding by Indication, Overadjustment |
| Performance/Attrition | Performance, Detection, Attrition, Differential Attrition |
| Behavioral/Expectancy | Hawthorne Effect, Confirmation Bias, Observer-Expectancy |
| Analysis/Design | Channeling, Surveillance, Data Dredging, Analysis Bias |
This is a shared conversation. Sign in to Orris to start your own chat.