Designing a Study You Can Defend
A secondary component in embedded mixed-methods research does not need to match the primary component in size. Learn how to determine an appropriate scope based on its research question, sampling requirements, analytical purpose, and contribution to integration.
Read Guide →
Designing a Study You Can Defend
An embedded component may grow beyond its original supporting role, but size alone does not make it a separate study. Learn how to evaluate methodological independence, integration, research questions, and reporting boundaries when an embedded component develops its own contribution.
Read Guide →
Designing a Study You Can Defend
Not every research question needs two groups, two treatments, or a before-and-after comparison. The comparison should follow from what you want to know, not from the assumption that stronger research always compares something.
Read Guide →
Designing a Study You Can Defend
A meaningful research comparison begins with identifying exactly what differs, between whom or what, and under which conditions. Learn how to translate a broad research question into a defensible comparison that supports the conclusions you intend to make.
Read Guide →
Designing a Study You Can Defend
Researchers often hear that comparison groups should differ only on the variable being investigated. This is a useful experimental ideal, but it is not a literal requirement for every valid study. Learn which differences matter, how randomization and design controls address alternative explanations, and when the intervention itself includes several inseparable components.
Read Guide →
Designing a Study You Can Defend
Perfectly comparable research groups are rarely achievable, particularly when studying people, institutions, or complex educational interventions. Learn why this limitation does not automatically invalidate a comparison, which differences matter most, and how to distinguish defensible inference from unsupported causal attribution.
Read Guide →
Designing a Study You Can Defend
Baseline differences are common when researchers compare existing classes, institutions, or self-selected participants. Learn how to anticipate these differences, define a defensible target comparison, select appropriate design and analytical strategies, and recognize when causal conclusions remain unsupported.
Read Guide →
Designing a Study You Can Defend
A comparison may address an important scientific question even when the available data cannot estimate its answer precisely. Learn how small samples, rare outcomes, limited overlap, clustering, and measurement uncertainty create statistical difficulties, and how to preserve scientific value without overstating the evidence.
Read Guide →
Designing a Study You Can Defend
Correlation, association, prediction, and causation answer different research questions. Learn how to match the claim you want to make with the evidence your study is actually designed to provide.
Read Guide →
Designing a Study You Can Defend
A causal question asks what would have happened under an alternative condition. That unobserved alternative is the counterfactual, and constructing a credible substitute for it is central to causal research.
Read Guide →
Designing a Study You Can Defend
Observational data do not automatically restrict researchers to non-causal questions. Causal inference may be possible when the causal contrast, design, assumptions, timing, confounding strategy, and analysis are explicitly aligned.
Read Guide →
Designing a Study You Can Defend
A proposed cause must precede its effect, but observing that one event happened before another does not establish causation. Learn why temporal precedence matters, how confounding and reverse causality can persist in longitudinal data, and which research designs provide stronger counterfactual evidence.
Read Guide →
Designing a Study You Can Defend
Regression adjustment, matching, and propensity-score methods can strengthen observational comparisons, but they do not automatically establish causation. Learn which biases adjustment can address, why unmeasured confounding remains consequential, and how to design observational studies around defensible causal assumptions.
Read Guide →
Designing a Study You Can Defend
A stronger outcome at higher exposure levels may support a causal explanation, but a dose-response gradient does not prove causation. Learn how exposure measurement, confounding, temporal order, nonlinear relationships, and study design affect the interpretation of dose-response evidence.
Read Guide →
Designing a Study You Can Defend
A causal study becomes more credible when its design directly addresses the most plausible explanation for the outcome other than the proposed cause. Learn how to identify rival explanations, prioritize threats to causal inference, choose design features that address them, and recognize when important uncertainty remains.
Read Guide →
Designing a Study You Can Defend
Baseline measurement should be guided by the research question, study design, and planned analysis rather than a desire to collect every available variable. Learn which measurements are essential, which are conditional, and how to avoid unnecessary data collection.
Read Guide →
Designing a Study You Can Defend
More follow-up measurements do not automatically produce better research. Learn how to determine the number of assessments needed to detect meaningful changes, examine outcome trajectories, and answer your research question without unnecessary participant burden.
Read Guide →
Designing a Study You Can Defend
An intervention may produce immediate improvements without lasting benefits, or its effects may emerge only after participants have had time to apply what they learned. Understand when both short-term and long-term outcomes should be measured and how to interpret differences between them.
Read Guide →
Designing a Study You Can Defend
A follow-up period is too short when it ends before the study can meaningfully observe the outcome or intervention effect it seeks to estimate. Learn how to recognize insufficient follow-up, distinguish early null findings from evidence of no effect, and establish an appropriate observation period.
Read Guide →
Designing a Study You Can Defend
Blinding is not always feasible, particularly in educational, behavioral, and complex intervention studies. Learn how to design credible unblinded research using safeguards that address specific sources of bias without pretending that every intervention can be concealed.
Read Guide →
Designing a Study You Can Defend
When participants are likely to share intervention materials or influence one another, contamination should shape the research design from the beginning. Learn how to select an appropriate randomization unit, preserve meaningful treatment contrasts, and plan for unavoidable spillover.
Read Guide →
Designing a Study You Can Defend
A completed session does not necessarily mean an intervention was delivered as intended. Learn how to assess intervention delivery through structured observations, component checklists, dose records, and quality ratings, while distinguishing faithful adaptation from implementation deviations.
Read Guide →
Designing a Study You Can Defend
An intervention may be delivered exactly as planned while participants engage with it inconsistently. Learn how participant adherence relates to intervention fidelity, why different frameworks define their relationship differently, and how to assess both without confusing delivery problems with participant behavior.
Read Guide →
Designing a Study You Can Defend
Explanatory and pragmatic trials differ primarily in the decisions they are designed to inform, not simply in how controlled or realistic they appear. Learn how their objectives influence eligibility, intervention delivery, comparison conditions, outcomes, and analysis, and why most trials occupy positions along a continuum.
Read Guide →
Designing a Study You Can Defend
Pragmatic trials can establish whether introducing an intervention improves outcomes under ordinary conditions, but identifying why it worked may be more difficult. Learn how implementation variation, context, and multiple causal pathways complicate explanation and how carefully designed process evaluation can help.
Read Guide →
