Manuel B. Garcia

Manuel B. Garcia serves as the Senior Director for Educational Technology and Digital Learning at FEU Institute of Technology, Manila, Philippines. Read More

Contact Info

1607, FEU Tech Building,
P. Paredes St, Sampaloc,
Manila, Philippines
mbgarcia@feutech.edu.ph

Follow Me

The Complete Library

All Research Guides

Browse every published guide in the Researcher's Field Guide, including the complete research journey and supplemental collections, listed alphabetically for quick reference.

Complete Library

Every published guide

Showing 2,951–3,000 of 5,065 guides

5,065 guides
PART 04 · SUBPART 19 · GUIDE 233

Understanding Natural Experiments

What Is a Natural Experiment?

A natural experiment uses exposure variation created by circumstances outside the researcher's control to investigate causal effects. Learn what qualifies as a natural experiment, why natural assignment does not automatically mean random assignment, and how researchers evaluate its credibility.

PART 04 · SUBPART 18 · GUIDE 232

Weak Quasi-Experiments vs. Strong Observational Studies

When Is a Weak Quasi-Experiment Less Informative Than a Strong Observational Study?

A quasi-experimental label does not guarantee stronger causal evidence than an observational study. Learn when weak comparisons undermine quasi-experiments, how rigorous observational designs can be more informative, and what researchers should prioritize when evaluating evidence.

PART 04 · SUBPART 18 · GUIDE 231

Choosing a Quasi-Experimental Design

How Do You Choose Among Quasi-Experimental Designs?

Choosing a quasi-experimental design requires more than matching a method to available data. Learn how treatment assignment, counterfactual assumptions, and identification opportunities determine which design can support a credible causal conclusion.

PART 04 · SUBPART 18 · GUIDE 230

Controlled vs. Uncontrolled Interrupted Time Series

When Is a Controlled Interrupted Time-Series Design Stronger Than an Uncontrolled Design?

Adding a comparison series can strengthen an interrupted time-series study, but only when the control provides a credible counterfactual. Learn when controlled designs offer stronger causal evidence, how to select suitable controls, and why an inappropriate comparison may introduce additional bias.

PART 04 · SUBPART 18 · GUIDE 228

When to Use Interrupted Time Series

When Is an Interrupted Time-Series Design Appropriate?

An interrupted time-series design examines whether an intervention changes an outcome's established pattern over time. Learn when this quasi-experimental approach is appropriate, how to interpret changes, and what can threaten causal inference.

PART 04 · SUBPART 18 · GUIDE 227

When to Use Instrumental Variables

When Is an Instrumental-Variable Design Appropriate for Causal Inference?

An instrumental-variable design may help estimate causal effects when treatment or exposure is confounded. Learn what makes an instrument credible, which assumptions matter, and why a statistical association alone is insufficient.

PART 04 · SUBPART 18 · GUIDE 226

Choosing a Valid Regression Discontinuity Cutoff

What Makes an Assignment Cutoff Suitable for Regression Discontinuity?

Not every eligibility threshold supports regression discontinuity. Learn how to evaluate whether a cutoff genuinely determines treatment, whether manipulation or competing policies threaten identification, and what evidence strengthens a local causal comparison.

PART 04 · SUBPART 18 · GUIDE 225

When to Use Regression Discontinuity

When Is Regression Discontinuity Appropriate?

Regression discontinuity can support credible causal inference when intervention assignment changes at a known threshold. Learn when the design is appropriate, how sharp and fuzzy designs differ, and why a cutoff alone is insufficient.

PART 04 · SUBPART 18 · GUIDE 224

Understanding the Parallel Trends Assumption

What Is the Parallel Trends Assumption in Difference-in-Differences?

Parallel trends is the central identifying assumption in many difference-in-differences designs. Learn what it actually requires, why baseline equality is unnecessary, how to assess its plausibility, and why passing a pre-trend test does not prove the assumption.

PART 04 · SUBPART 18 · GUIDE 223

When to Use Difference-in-Differences

When Is Difference-in-Differences an Appropriate Design Strategy?

Difference-in-differences can estimate intervention effects by comparing outcome changes between treated and comparison groups. Learn when this strategy is appropriate, which assumptions matter, and why having before-and-after data is not enough.

PART 04 · SUBPART 18 · GUIDE 222

Choosing a Credible Comparison Group

What Makes a Comparison Group Credible in a Quasi-Experiment?

A comparison group is not credible merely because it resembles the intervention group. Learn how assignment mechanisms, untreated outcome trajectories, eligibility, and exposure differences determine whether a comparison can support causal inference.

PART 04 · SUBPART 18 · GUIDE 221

Causal Inference in Quasi-Experiments

Can a Quasi-Experiment Support Causal Inference?

Quasi-experiments can support causal conclusions without random assignment, but not simply because they compare treated and untreated groups. Learn how credible counterfactuals, identification assumptions, and sensitivity analyses determine the strength and limits of causal inference.

PART 04 · SUBPART 18 · GUIDE 220

Quasi-Experimental vs. Observational Research

What Makes a Study Quasi-Experimental Rather Than Observational?

Quasi-experimental and observational studies can both involve nonrandomized exposure, making their boundaries less obvious than textbooks suggest. Learn how intervention assignment, causal identification, comparison strategies, and disciplinary conventions distinguish the approaches.

PART 04 · SUBPART 18 · GUIDE 219

When to Use Quasi-Experimental Research

When Should You Use a Quasi-Experimental Design?

Quasi-experimental designs can evaluate interventions when random assignment is unavailable or inappropriate. Learn when these designs are justified, which causal assumptions matter, and how to distinguish a credible quasi-experiment from a weak before-and-after comparison.

PART 04 · SUBPART 17 · GUIDE 218

Time-Related Confounding in Rollout Designs

When Does Time-Related Confounding Threaten the Validity of a Rollout Design?

Interventions introduced at different times can appear effective because outcomes were already changing or other events occurred during implementation. Learn when time-related confounding threatens rollout evaluations, how it differs from ordinary secular trends, and what researchers can do to strengthen causal inference.

PART 04 · SUBPART 17 · GUIDE 217

Research Using Naturally Staged Rollouts

When Can a Naturally Staged Intervention Rollout Be Used for Research?

A naturally staged intervention rollout can create an opportunity to evaluate a policy or program without controlling its implementation. Learn when staggered adoption supports credible research, how to identify comparison groups, and why implementation timing and pre-intervention trends matter.

PART 04 · SUBPART 17 · GUIDE 216

When Rollout Logistics Shape Research Design

Can Intervention Rollout Logistics Determine the Research Design?

Implementation capacity, staffing, infrastructure, and institutional schedules can influence which research designs are feasible. Learn when rollout logistics legitimately shape a study and why administrative convenience should not be mistaken for methodological justification.

PART 04 · SUBPART 17 · GUIDE 215

Planning Stepped-Wedge Steps and Periods

How Many Steps and Measurement Periods Does a Stepped-Wedge Design Need?

There is no universal number of steps or measurement periods for a stepped-wedge trial. Learn how crossover sequences, cluster numbers, implementation delays, temporal correlation, and outcome measurement determine an appropriate design.

PART 04 · SUBPART 17 · GUIDE 214

Secular Time Trends in Stepped-Wedge Trials

How Do Secular Time Trends Affect Stepped-Wedge Studies?

Stepped-wedge trials introduce interventions progressively, making treatment exposure closely related to calendar time. Learn how secular trends can distort estimated effects, when standard period adjustment may be insufficient, and how to design more credible temporal comparisons.

PART 04 · SUBPART 17 · GUIDE 213

Stepped-Wedge vs. Parallel Cluster Trials

When Is a Stepped-Wedge Design Better Than a Parallel Cluster-Randomized Trial?

Stepped-wedge and parallel cluster-randomized trials can evaluate group-level interventions, but they differ in allocation, timing, analysis, and implementation demands. Learn when phased randomized rollout provides a genuine advantage and when a simpler parallel trial may be preferable.

PART 04 · SUBPART 17 · GUIDE 212

Are Stepped-Wedge Designs More Ethical?

Does Giving Every Cluster the Intervention Make a Stepped-Wedge Design More Ethical?

Giving every cluster eventual access to an intervention may appear fairer than withholding it from a comparison group. Yet stepped-wedge trials raise distinct ethical questions about delayed access, uncertain benefits, consent, and exposure to potentially harmful interventions.

PART 04 · SUBPART 17 · GUIDE 211

Understanding Stepped-Wedge Designs

What Is a Stepped-Wedge Design, and When Is It Useful?

A stepped-wedge design introduces an intervention to different clusters at different times, often using randomized rollout sequences. Understand how the design works, when it offers advantages over a parallel trial, and why time-related confounding requires particular attention.

PART 04 · SUBPART 16 · GUIDE 210

Individual Randomization Despite Contamination

When Is Individual Randomization Preferable Despite Some Contamination Risk?

Some contamination risk does not automatically justify randomizing entire groups. Learn when individual randomization offers a more efficient and credible design, how to manage spillover, and when contamination becomes serious enough to reconsider the allocation unit.

PART 04 · SUBPART 16 · GUIDE 209

Handling Unequal Cluster Sizes

How Do You Handle Unequal Cluster Sizes in a Cluster-Randomized Trial?

Unequal cluster sizes are common in school, hospital, and community trials. Understand how size variation affects statistical efficiency, treatment-effect weighting, and sample-size planning, and how to choose an appropriate analytical strategy.

PART 04 · SUBPART 16 · GUIDE 208

Cluster Allocation and Recruitment Bias

Can Knowledge of Cluster Allocation Introduce Recruitment Bias?

Even correctly randomized clusters can produce biased participant comparisons when recruiters know the assigned intervention. Understand how allocation knowledge influences eligibility, enrollment, and consent, and which safeguards help preserve causal validity.

PART 04 · SUBPART 16 · GUIDE 207

Recruitment After Cluster Randomization

Can Participants Be Recruited After Clusters Are Randomized?

Participants can be recruited after clusters have been randomized, but knowledge of treatment allocation may influence who enters the trial. Learn when post-randomization recruitment is necessary, what biases it can introduce, and how to design defensible recruitment procedures.

PART 04 · SUBPART 16 · GUIDE 206

Matching vs. Stratifying Clusters

Should Clusters Be Matched or Stratified Before Randomization?

Matching and stratification can improve baseline balance in cluster-randomized trials, particularly when relatively few groups are available. Learn how these approaches differ, when each may be useful, and why matching does not automatically improve statistical power.

PART 04 · SUBPART 16 · GUIDE 205

Many Participants, Few Clusters

Is Having Many Participants but Very Few Clusters Enough in a Cluster-Randomized Trial?

A cluster-randomized trial can involve thousands of participants yet have too few independently randomized groups for reliable inference. Learn why increasing participant numbers cannot fully compensate for a small cluster count and what researchers can do when additional clusters are unavailable.

PART 04 · SUBPART 16 · GUIDE 204

How Many Clusters Do You Need?

How Many Clusters Do You Need in a Cluster-Randomized Trial?

There is no universal minimum number of clusters for a cluster-randomized trial. Learn how to calculate a defensible number, why very few clusters create statistical problems, and how to balance cluster recruitment against participant enrollment.

PART 04 · SUBPART 16 · GUIDE 203

Why Cluster Trials Need Larger Samples

Why Does Cluster Randomization Usually Require a Larger Sample?

Cluster-randomized trials often need larger samples because participants within the same cluster provide correlated rather than fully independent information. Learn how intracluster correlation, cluster size, and the number of randomized groups affect statistical power and sample-size planning.

PART 04 · SUBPART 16 · GUIDE 202

Costs and Limitations of Cluster Randomization

What Are the Costs and Limitations of Cluster Randomization?

Cluster randomization can accommodate group-level interventions, but it introduces statistical inefficiency, recruitment risks, and operational demands. Understand its principal costs and limitations and how to assess whether the design is justified.

PART 04 · SUBPART 16 · GUIDE 201

Contamination and Cluster Randomization

When Does Contamination Make Cluster Randomization Preferable?

Contamination can undermine the contrast between intervention and comparison groups, but its mere possibility does not justify randomizing entire groups. Learn how to assess contamination mechanisms, their likely consequences, and whether cluster randomization offers a defensible solution.

PART 04 · SUBPART 16 · GUIDE 200

What Counts as a Cluster?

What Counts as a Cluster in a Cluster-Randomized Trial?

A cluster is the group or unit assigned to an intervention condition in a cluster-randomized trial. Learn what qualifies as a cluster, how to define its boundaries, and why choosing the correct unit matters for valid statistical inference.

PART 04 · SUBPART 16 · GUIDE 199

When to Use Cluster Randomization

When Should Groups Rather Than Individuals Be Randomized?

Randomizing groups instead of individuals can be appropriate when an intervention operates at the group level or individual allocation would compromise implementation. Learn when cluster randomization is justified and what methodological costs accompany that decision.

PART 04 · SUBPART 15 · GUIDE 198

When to Use Adaptive Multi-Arm Trials

When Is an Adaptive Multi-Arm Trial Worth the Added Complexity?

Adaptive multi-arm trials can modify prespecified aspects of a study using accumulating evidence, potentially conserving resources and reducing exposure to unpromising interventions. Learn when these advantages justify the additional statistical, operational, and governance requirements, and when a conventional multi-arm trial remains the better choice.

PART 04 · SUBPART 15 · GUIDE 197

Dropping Ineffective Trial Arms

When Should an Ineffective Treatment Arm Be Dropped During a Multi-Arm Trial?

An ineffective intervention arm may be discontinued during a multi-arm trial, but the decision should not rest on an ordinary nonsignificant result. Learn how prespecified futility rules, interim analyses, independent monitoring, and adaptive designs support defensible decisions while protecting participants and statistical inference.

PART 04 · SUBPART 15 · GUIDE 196

Unequal Sample Sizes Across Trial Arms

Does Every Treatment Arm in a Multi-Arm Trial Need the Same Sample Size?

Randomized trials do not always require equal numbers of participants in every arm. Unequal allocation may improve precision, reduce costs, or address ethical and operational constraints, but it can also reduce statistical power. Learn when unequal allocation is justified and how to evaluate its consequences.

PART 04 · SUBPART 15 · GUIDE 195

Multi-Arm Trial Recruitment Feasibility

How Many Treatment Arms Can a Trial Support Before Recruitment Becomes Unrealistic?

Adding treatment arms increases the comparisons a trial can investigate, but recruitment capacity may not increase with them. Learn how to determine a feasible number of arms, estimate enrollment requirements, evaluate recruitment timelines, and recognize when a simpler design will produce stronger evidence.

PART 04 · SUBPART 15 · GUIDE 194

Multi-Arm vs. Separate Two-Arm Trials

When Is a Multi-Arm Trial Better Than Several Separate Two-Arm Trials?

A multi-arm trial can compare several interventions within one coordinated study, often reducing duplicated recruitment and allowing treatments to share a control group. Learn when this approach is more efficient, how multiplicity affects inference, and when separate two-arm trials remain preferable.

PART 04 · SUBPART 15 · GUIDE 193

Designing for Intervention Interactions

How Should Researchers Design a Factorial Trial When Interventions May Strongly Interact?

When interventions may strongly interact, a study designed only for average treatment effects may conceal important differences between combinations. Learn how to define interaction hypotheses, plan adequate sample size, select appropriate comparisons, and interpret combined intervention effects.

PART 04 · SUBPART 15 · GUIDE 192

The No-Interaction Assumption in Factorial Designs

When Can a Factorial Trial Assume That Interventions Do Not Interact?

Factorial designs do not inherently assume that interventions operate independently. However, some efficiency-oriented factorial trials rely on negligible interactions when planning and interpreting marginal treatment effects. Learn when this assumption matters, how it differs from statistical independence, and what researchers should do when interactions remain plausible.

PART 04 · SUBPART 15 · GUIDE 191

Testing Interactions in Factorial Designs

Can a Factorial Design Test Interactions Between Interventions?

Factorial designs can test whether the effect of one intervention changes when another is present. Learn how interaction contrasts are calculated, why the outcome scale matters, and how to distinguish meaningful interactions from statistical uncertainty.

PART 04 · SUBPART 15 · GUIDE 190

Factorial Design vs. Separate Trials

What Can a Factorial Design Reveal That Separate Trials Cannot?

Separate trials can establish whether individual interventions work, but they may not reveal how those interventions perform together. Learn which additional questions a factorial design can answer, when separate trials remain appropriate, and why interaction claims require careful planning.

PART 04 · SUBPART 15 · GUIDE 189

When to Use a Factorial Design

When Should You Use a Factorial Design?

Factorial designs allow researchers to investigate multiple experimental factors within one study. Learn when they are useful for evaluating interventions, how interactions affect interpretation, and why research objectives must determine the allocation, analysis, and sample size.

PART 04 · SUBPART 14 · GUIDE 188

Parallel-Group vs. Crossover Design

When Is a Parallel-Group Design Preferable to a Crossover Design?

Parallel-group and crossover trials answer treatment questions through different comparison structures. Learn when parallel groups provide more defensible evidence, why crossover designs may improve precision, and how treatment persistence, time-related change, and participant burden should influence the choice.

PART 04 · SUBPART 14 · GUIDE 187

Irreversible Interventions and Crossover Designs

Can an Irreversible Intervention Be Studied Using a Crossover Design?

Irreversible interventions create fundamental problems for conventional crossover trials because participants cannot return to their original state between treatments. Learn when this prevents valid comparisons, what exceptions require careful interpretation, and which alternative designs may better answer the research question.

PART 04 · SUBPART 14 · GUIDE 186

Determining Washout Period Duration

How Long Should a Washout Period Be in a Crossover Trial?

A washout period should be long enough to make residual treatment effects sufficiently negligible for the intended comparison. Learn how pharmacokinetics, biological recovery, behavioral persistence, and trial feasibility influence its duration, including why no universal washout rule applies.

PART 04 · SUBPART 14 · GUIDE 185

Carryover Effects in Crossover Trials

What Is a Carryover Effect in a Crossover Trial?

A carryover effect occurs when an earlier treatment continues to influence outcomes during a later treatment period. Learn how carryover threatens crossover comparisons, why it can be difficult to detect statistically, and which design decisions can reduce its influence.

PART 04 · SUBPART 14 · GUIDE 184

When Crossover Trials Are Inappropriate

When Is a Crossover Trial Inappropriate?

Crossover trials can improve statistical precision, but their assumptions are not always defensible. Learn when persistent effects, unstable conditions, ethical constraints, or incomplete treatment sequences make a crossover design unsuitable.