Designing a Study You Can Defend
Existing registries and cohorts can provide valuable research data without requiring investigators to recruit an entirely new sample. Their suitability depends on whether the population, measurements, follow-up, and data quality can support the intended research question.
Read Guide →
Designing a Study You Can Defend
A registry may contain extensive records without adequately representing the population you want to study. Learn how to evaluate population alignment, identify meaningful coverage differences, and determine how broadly your findings can reasonably apply.
Read Guide →
Designing a Study You Can Defend
A registry can support a nested case-control study when it defines a suitable source cohort and provides reliable event timing and exposure information. Learn how to construct risk sets, select controls, and analyze the resulting data without losing the logic of the underlying cohort.
Read Guide →
Designing a Study You Can Defend
Registry variables may be introduced, discontinued, or collected only during certain periods. Learn how to distinguish structural missingness from ordinary missing data, establish valid analysis periods, and avoid misleading comparisons.
Read Guide →
Designing a Study You Can Defend
The unit of analysis determines what your study is actually analyzing and what your conclusions concern. Learn how to identify the appropriate unit, distinguish it from the unit of observation, and avoid errors involving clustered or repeated data.
Read Guide →
Designing a Study You Can Defend
Different research questions require different analytical units, even when they use similar data. Compare individual, household, classroom, institutional, community, and country-level analysis to determine which level supports your intended conclusions.
Read Guide →
Designing a Study You Can Defend
The entities from which researchers collect observations are not always the entities they analyze. Learn how to distinguish units of observation from units of analysis, recognize legitimate mismatches, and choose methods that support the intended conclusions.
Read Guide →
Designing a Study You Can Defend
Misidentifying the unit of analysis can produce misleading sample sizes, inappropriate statistical tests, and conclusions at the wrong level. Learn how to recognize these errors, assess their consequences, and determine whether an existing analysis can be corrected.
Read Guide →
Designing a Study You Can Defend
A dataset with thousands of records may contain far fewer independent participants or experimental units. Learn how to distinguish observations from independent replication, calculate illustrative design effects, and prevent misleading sample size and statistical precision.
Read Guide →
Designing a Study You Can Defend
Researchers may recruit schools, hospitals, households, or organizations while analyzing outcomes from individuals within them. Learn when recruitment and analytical units can differ, how this affects selection and clustering, and what must be reported.
Read Guide →
Designing a Study You Can Defend
Researchers can randomize classrooms, hospitals, or organizations while measuring outcomes from individuals within them. Learn how to distinguish randomization and measurement units, plan appropriate analyses, and avoid misleading treatment-effect estimates.
Read Guide →
Designing a Study You Can Defend
Recruiting individuals through selected schools, hospitals, organizations, or communities can make research more feasible, but it also affects population coverage, selection probabilities, statistical dependence, and precision. Learn how to design and analyze these studies without confusing the number of participants with the number of independent groups.
Read Guide →
Designing a Study You Can Defend
Assigning treatment to classrooms, hospitals, or communities while measuring individual outcomes creates a clustered intervention design. Learn how this affects treatment-effect estimation, statistical independence, sample size, recruitment bias, and causal interpretation.
Read Guide →
Designing a Study You Can Defend
A study with hundreds of participants may still have too few independent clusters to support reliable inference. Learn how to determine the appropriate number of participants and clusters, calculate illustrative design effects, and plan sample size according to the research question and assignment structure.
Read Guide →
Designing a Study You Can Defend
Pseudoreplication can arise when researchers treat multiple measurements or participants within one treatment-assigned group as independent experimental replications. Learn how to identify genuine treatment replication, distinguish subsamples from experimental units, and determine whether a mismatched analysis can be corrected.
Read Guide →
Designing a Study You Can Defend
Interventions may operate at the classroom, school, hospital, or community level while outcomes are measured on individuals. Learn how to align intervention delivery, treatment assignment, outcome measurement, sampling, and statistical inference within a defensible research design.
Read Guide →
Designing a Study You Can Defend
A defensible sample begins with a clearly defined population. Learn how to move from your research question to the people or units you actually recruit without claiming more than your design can support.
Read Guide →
Designing a Study You Can Defend
Your data usually come from a sample, but your research question may concern a much larger population. Understanding that distinction is essential for choosing participants and interpreting what your findings actually mean.
Read Guide →
Designing a Study You Can Defend
A well-defined target population establishes who or what your research is intended to describe or explain. Learn how to specify population boundaries, align them with your research question, distinguish them from recruitment restrictions, and avoid unsupported generalization.
Read Guide →
Designing a Study You Can Defend
A target population that is too broad may exceed what the study can meaningfully investigate, while one that is unnecessarily narrow may limit relevance and generalizability. Learn how to establish defensible population boundaries using the research question, estimand, relevant heterogeneity, and intended application of the findings.
Read Guide →
Designing a Study You Can Defend
Choosing study sites before defining the target population can allow convenience and institutional access to determine the scope of research conclusions. Learn when population definition should precede site selection, how to select appropriate research settings, and when site-first research designs are justified.
Read Guide →
Designing a Study You Can Defend
A research population exists within a temporal context, but the dates of data collection do not necessarily define that population. Learn when time should determine eligibility, how to distinguish population reference periods from observation periods, and why temporal boundaries matter for valid research conclusions.
Read Guide →
Designing a Study You Can Defend
Restricting a research population can improve conceptual clarity, measurement comparability, and intervention relevance, but it may also limit who the findings apply to. Learn when narrowing the population is justified, why homogeneity does not automatically guarantee internal validity, and how to balance focused research questions with broader applicability.
Read Guide →
Designing a Study You Can Defend
Inclusion criteria should identify the participants or other units necessary to answer your research question, not merely describe who is easiest to recruit. Learn how to translate a target population into clear, operational eligibility rules, avoid unnecessary restrictions, and ensure that participant selection supports the intended inference.
Read Guide →
Designing a Study You Can Defend
Inclusion criteria should be determined by more than the characteristics of participants who are easiest to recruit. Learn how the research question, target population, study design, intervention requirements, measurement validity, ethics, and generalizability should inform eligibility decisions.
Read Guide →