Designing a Study You Can Defend
Restriction can reduce confounding by limiting variation in selected characteristics, but it also changes the population represented by a study. Learn when restriction improves causal comparability, how it affects generalizability and precision, and why restricting on the wrong variable may introduce bias.
Read Guide →
Designing a Study You Can Defend
Cases and controls may remember or report past exposures differently because of their outcome status, beliefs, or experiences. Learn how differential recall can distort case-control associations, why its direction is not predictable without assumptions, and how researchers can design more credible retrospective exposure assessments.
Read Guide →
Designing a Study You Can Defend
Interviewers can influence research findings through differences in question wording, probing, tone, and interpretation across study groups. Learn when these differences create information bias, how they affect exposure and outcome measurement, and which design safeguards improve comparability.
Read Guide →
Designing a Study You Can Defend
Differences in device calibration can make identical underlying quantities produce different recorded measurements. Learn how systematic error, calibration drift, and device allocation can distort research comparisons, and how researchers can establish measurement comparability through calibration, verification, and sensitivity analysis.
Read Guide →
Designing a Study You Can Defend
Blinding reduces the influence of knowing study-group status, while standardization addresses differences in how measurements are collected. Learn when procedural consistency deserves priority, when assessor blinding is essential, and how to combine both safeguards without confusing their purposes.
Read Guide →
Designing a Study You Can Defend
A complete dataset is not necessarily an accurate one. Learn how to identify critical research variables, design meaningful validation checks, distinguish data errors from legitimate observations, and document corrections without compromising research integrity.
Read Guide →
Designing a Study You Can Defend
Statistical adjustment can reduce confounding when the right variables are measured and modeled appropriately, but it cannot guarantee an unbiased estimate. Unmeasured confounders, measurement error, model misspecification, and inappropriate adjustment can leave or even introduce bias.
Read Guide →
Designing a Study You Can Defend
Identifying potential confounders before data collection requires more than listing variables associated with an outcome. Learn how to distinguish plausible common causes from mediators, colliders, and other variables that may not belong in an adjustment strategy.
Read Guide →
Designing a Study You Can Defend
Restricting eligibility can reduce confounding by limiting variation in selected participant characteristics, but it also narrows the population studied. Understand when restriction is appropriate, when it may introduce problems, and how to justify eligibility criteria.
Read Guide →
Designing a Study You Can Defend
Matching participants can improve comparability between study groups on selected characteristics, but it does not automatically eliminate confounding. Understand when matching is appropriate, which variables to match, and how to evaluate the resulting sample and analysis.
Read Guide →
Designing a Study You Can Defend
Stratified sampling and stratified randomization both divide participants into subgroups, but they serve different methodological purposes. Learn how each affects representation, treatment allocation, precision, and confounding control.
Read Guide →
Designing a Study You Can Defend
Collecting extra variables in case they become useful for statistical adjustment may seem cautious, but more data do not automatically produce more credible results. Learn how to decide which variables deserve measurement before a study begins.
Read Guide →
Designing a Study You Can Defend
Failing to measure important confounders can leave observational findings vulnerable to alternative causal explanations, even after statistical adjustment. Learn how to recognize insufficient measurement, assess its consequences, and strengthen research design before data collection.
Read Guide →
Designing a Study You Can Defend
Adjusting for a mediator can remove part of the causal effect you intend to estimate, while conditioning on a collider can create a misleading association. Understand the different mechanisms, practical examples, and how to avoid inappropriate adjustment.
Read Guide →
Designing a Study You Can Defend
Confounding control begins before statistical analysis. Learn how to develop a defensible strategy that connects your causal question, participant selection, confounder measurement, analytical methods, and sensitivity analyses before data collection starts.
Read Guide →
Designing a Study You Can Defend
Some research variables are more likely to be missing because of their sensitivity, timing, complexity, or collection procedures. Learn how to identify vulnerable measurements before data collection and prioritize practical prevention without confusing missingness risk with statistical importance.
Read Guide →
Designing a Study You Can Defend
Missing data may occur differently across study groups because of participant characteristics, intervention demands, or outcome-related processes. Learn how to distinguish unequal missingness rates from biased comparisons and how to investigate and address group-specific missing-data risks.
Read Guide →
Designing a Study You Can Defend
Changing when and how often researchers collect measurements can reduce preventable missingness, but poorly chosen schedules may alter the outcomes being measured. Learn how to improve follow-up timing while preserving scientific validity and comparability.
Read Guide →
Designing a Study You Can Defend
Backup data sources can preserve critical outcomes when planned assessments are unavailable, but alternative records must measure the intended outcome accurately. Learn when backup sources are justified, how to establish source equivalence, and what to specify before data collection.
Read Guide →
Designing a Study You Can Defend
There is no universal percentage of missing data that makes a study invalid. The consequences depend on which information is missing, why it is unavailable, and whether the remaining observations can support the primary analysis. Learn how to assess missingness using precision, identification assumptions, and sensitivity analyses rather than arbitrary cutoffs.
Read Guide →
Designing a Study You Can Defend
Case-control studies can investigate uncommon outcomes without following enormous populations to accumulate cases. Learn when this efficiency advantage matters, how case and control sampling affects validity, and why an efficient design still requires careful attention to the source population, exposure measurement, and causal assumptions.
Read Guide →
Designing a Study You Can Defend
Cohort studies are often well suited to rare exposures because researchers can identify exposed individuals directly and investigate their subsequent outcomes. Learn when this approach is preferable, how to construct an appropriate comparison group, and why exposure-based recruitment still requires careful attention to confounding, timing, and selection bias.
Read Guide →
Designing a Study You Can Defend
Recruiting across multiple research sites may make studies of rare cases feasible, but additional sites can introduce differences in populations, outcome definitions, and measurement procedures. Learn when multisite recruitment is justified, how to harmonize data collection, and how to evaluate whether the additional cases strengthen the intended inference.
Read Guide →
Designing a Study You Can Defend
Historical data can increase the number of observations available for research, especially when current recruitment is limited or outcomes are rare. Learn when older records can be combined with new data, how to evaluate changes over time, and why a larger sample may not provide a more valid comparison.
Read Guide →
Designing a Study You Can Defend
Research registries can make rare-outcome studies feasible by identifying cases across institutions, geographic areas, and extended periods. However, registry size alone does not establish validity. Learn how to evaluate coverage, case definitions, exposure information, selection bias, and whether a registry can support the intended research design.
Read Guide →
