BCBR • High-Yield 100 MCQs

Legitimate authorship requires meaningful contribution and responsibility/accountability according to accepted authorship standards.

Because participants are selected by outcome status, risks/incidence usually cannot be estimated directly; the odds ratio is the standard association measure.

Randomisation reduces systematic baseline differences between intervention groups on average; it does not itself ensure blinding or complete follow-up.

Data collection tools should derive from the objectives and variables needed to answer the research question.

Voluntary participation includes the right to withdraw, while handling of already collected data follows consent, protocol, ethics and applicable rules.

The primary objective drives design, sample size, data collection and the principal analysis.

Blinding assessors reduces measurement/ascertainment bias influenced by knowledge of intervention or exposure.

SD describes variability among observations; standard error describes precision of an estimated mean.

Prespecification reduces selective outcome reporting and links sample size and analysis to the main objective.

Participant safety comes first; serious events require management and reporting under the protocol and applicable regulatory/ethics framework.

Predictive values vary with pre-test probability/prevalence as well as test sensitivity/specificity.

Focused questions guide objectives, design, variables, sample size and analysis.

A cross-sectional study estimates prevalence by measuring exposure/outcome status in a defined population at a point or short period.

Incidence concerns new events among persons at risk over a specified time; prevalence describes existing disease burden.

Concealment prevents recruiters from predicting the next assignment and selectively enrolling participants.

A confounder is related to both exposure and outcome but is not an intermediate step in the causal pathway.

Type I error is a false-positive statistical conclusion; its probability is controlled by alpha.

Stratified sampling deliberately samples within defined strata, often to improve representation or precision.

Higher power generally requires a larger sample for the same effect size, alpha and variability.

The p value is computed under the null/model assumptions; it is not the posterior probability that the null is true and does not establish clinical importance.

Piloting checks comprehension, flow, feasibility, coding and field procedures before full deployment.

Consent is a process requiring relevant disclosure, understanding and voluntary choice; a signature alone is insufficient.

Plagiarism is misappropriation of others' work or ideas without proper attribution.

A cohort design classifies exposure before outcome follow-up, supporting incidence estimation and temporality.

Intention-to-treat preserves the randomized comparison by retaining participants in their assigned groups, subject to appropriate handling of missing outcomes.

Recall bias is a form of information bias particularly relevant when past exposures are remembered differently by outcome groups.

Differential or prognosis-related attrition can distort group comparisons and introduce selection bias.

For common estimators, standard error decreases as sample size increases, reflecting greater precision.

Probability sampling uses a defined random selection mechanism rather than convenience or investigator choice.

Convenience samples are selected for ease rather than probability, which can limit representativeness.

Greater precision requires more observations, all else equal.

Focus groups generate qualitative data about experiences, views, meanings and social interaction.

Purposive sampling intentionally selects participants likely to provide relevant, diverse or rich perspectives.

Thematic analysis systematically codes qualitative material and develops patterns of meaning.

Vulnerability calls for proportionate protections, not blanket exclusion when research is ethically justified.

Payments/benefits must be structured so they do not improperly compromise voluntariness, especially in vulnerable settings.

Width reflects statistical precision under the model; narrower intervals generally mean more precise estimation.

For a ratio measure, 1 is the null value; inclusion generally corresponds to p≥0.05 for the matching two-sided test.

Following outcome-free participants over time permits direct incidence estimation.

Undisclosed redundant publication can double-count evidence and breach publication ethics.

Effect modification/interaction can be scientifically meaningful heterogeneity, whereas confounding is distortion to control.

Matching can control selected confounding factors but must be accounted for in analysis and may introduce overmatching concerns.

Median and IQR are resistant to extreme values and better represent a skewed distribution.

A confidence interval expresses uncertainty/precision around an estimate under the analysis assumptions; it is not the probability that a fixed parameter lies inside a realised interval.

Power is 1−beta: the probability of detecting a specified effect when that effect truly exists under the model.

Blood group consists of unordered categories, making it nominal.

Ordered categories such as stages are ordinal because rank matters but intervals need not be equal.

Coded identifiers with controlled access to the re-identification key reduce unnecessary exposure of direct identifiers.

Justice addresses fairness in participant selection and distribution of research burdens and benefits.

Fabrication is making up data or results and recording/reporting them as if real.

The null typically specifies no effect/difference/association for the parameter being tested.

Failure to reach a significance threshold may reflect imprecision; effect size and confidence interval should be examined.

Cluster designs sample groups, often for logistical efficiency, and analysis/design must account for within-cluster similarity.

An independent t test compares two independent group means under its assumptions.

A data dictionary documents how variables are represented and interpreted, supporting consistent data management/analysis.

Independent entry/comparison or validation rules can detect transcription and coding errors.

Prespecifying analysis improves coherence and reduces opportunistic analytical choices.

Background/rationale synthesises existing knowledge and identifies the gap motivating the study.

Prospective registration improves transparency and helps identify selective reporting or unpublished trials.

Independent ethics review evaluates scientific/ethical acceptability with focus on participant rights, safety and welfare.

Coding/pseudonymisation can remain reversible via a key; de-identification seeks to make re-identification substantially less feasible.

Secure, tested backups are part of data integrity and continuity planning.

Appraisal considers validity/risk of bias before interpreting effects and applicability.

Prevalence reflects both occurrence and how long individuals remain in the diseased state.

Simultaneous measurement can make it unclear whether exposure preceded outcome.

Sensitivity is P(test positive | disease).

Specificity is P(test negative | no disease).

Transparent assessment and appropriate methods are needed because missingness can reduce precision and introduce bias.

Testing many hypotheses without appropriate control raises the probability of at least one chance-positive result.

Boolean AND intersects concepts; OR usually broadens within a concept set.

OR captures records containing any of the alternative terms/synonyms.

A scientifically interesting question must also be practically achievable with available resources and access.

Risk/cumulative incidence begins with a disease-free population at risk for the event.

Incidence rates use person-time to accommodate varying observation durations.

Transparent documentation preserves integrity and permits appropriate safety and analytical assessment.

Version/date tracking prevents use of outdated forms/protocols and supports auditability.

A codebook/data dictionary makes datasets interpretable and reproducible.

Validation checks flag values outside allowable ranges or inconsistent combinations for review.

External validity concerns whether findings apply to other populations, settings or circumstances.

A histogram displays the frequency distribution and shape of continuous data grouped into intervals.

Large samples can detect very small effects; clinical importance depends on magnitude, harms, costs and context.

The chi-square test commonly assesses association between categorical variables in contingency tables when expected counts are adequate.

The paired t test uses within-person differences for paired continuous observations.

Pearson r measures linear association; a near-zero value does not exclude nonlinear relationships.

Logistic regression models the log odds of a binary outcome and can adjust for multiple predictors.

Predefined inclusion/exclusion criteria improve consistency, reproducibility and protection against post hoc selection.

Gantt charts help schedule tasks, milestones and dependencies across a project timeline.

No absolute difference corresponds to risk difference zero.

Ecological studies analyse aggregate/group-level exposure and outcome measures.

Relationships observed across groups need not represent individual-level relationships.

Prevalence = existing cases ÷ population assessed = 80/400 = 0.20 = 20%.

Cumulative incidence = new cases ÷ persons initially at risk = 25/500 = 5%.

Risk exposed = 0.20; risk unexposed = 0.05; RR = 0.20/0.05 = 4.

OR = ad/bc = (60×70)/(40×30) = 4200/1200 = 3.5.

n = 4×50×50/10² = 10000/100 = 100. This is the common approximate prevalence sample-size calculation.

Sensitivity = true positives ÷ all with disease = 90/100 = 90%.

Specificity = true negatives ÷ all without disease = 180/200 = 90%.

For a ratio, the null is 1. The entire 95% CI is below 1, so the corresponding two-sided test is conventionally significant; this does not prove absence of bias or an individual-level benefit.

For a difference measure, 0 is the null. Because the interval crosses 0, the data are compatible with no difference at the conventional two-sided 5% level.

Sampling addresses selection from a population; randomisation addresses allocation to intervention groups. This distinction is a classic research-methodology exam trap.