# Global Health Research in Practice

> Companion website for *Global Health Research in Practice*, a textbook by Eric Green (published as E. Philip Green) on research design and causal inference for global health. The site hosts the canonical online presence for the book and a growing set of code-driven companion notes that extend specific chapters with real-world worked examples.

The book teaches practical methods for designing and interpreting health research in diverse, resource-constrained settings: causal reasoning, experimental and quasi-experimental design, statistical inference, external validity, and the feasibility and implementation realities of fieldwork. The companion notes connect a single methodological concept to a concrete, current case (a trial, a policy, a regulatory decision) with reproducible R code and figures.

Author: Eric Green (E. Philip Green), global health methods scholar.

## Key resources

- [Book home](https://ghrbook.com/): overview of the book and where to read or buy it.
- [About the book](https://ghrbook.com/about): synopsis and table of contents.
- [About the author](https://ghrbook.com/author): background and contact.
- [Read online](https://read.ghrbook.com): the full book as an online reader.
- [Course](https://learn.ghrbook.com/): the companion course.
- [Resources](https://ghrbook.com/resources): index of all companion notes.
- [Chapter videos](https://ghrbook.com/videos): short async lecture videos, with slides and transcripts.
- [Video manifest](https://ghrbook.com/videos.json): machine-readable index of every video, its slides, and its transcript.
- [Research Notes podcast](https://ghrbook.com/podcast): interviews tied to the notes.

## Companion notes

- [What Would It Take to Stop This Outbreak?](https://ghrbook.com/notes/bundibugyo-outbreak-model.html): An Interactive Branching-Process Model of the 2026 Bundibugyo Epidemic. Reproduce the CDC's 2026 Bundibugyo outbreak projections in your browser, then change the assumptions behind them. An interactive branching-process model that shows how modeling studies inform a public health response: every conclusion is conditional on a stated mechanism, and isolation coverage near 70% is where containment tips. (Ch. 14 Other Designs; topics: Outbreak Modeling, Branching Process, Reproduction Number, Uncertainty) (Research Notes podcast episode with LCDR Eric Mooring, U.S. Centers for Disease Control and Prevention)
- [Inside the Bundibugyo Outbreak Model](https://ghrbook.com/notes/bundibugyo-model-technical.html): Parameters, Code, and Validation Against the CDC's Published Figures. The technical companion to the interactive Bundibugyo model: every parameter from the MMWR prior box, the core of the code, and a side-by-side validation showing the reimplementation reproduces the report's three supplementary figures and inferred spillover dates. (Ch. 14 Other Designs; topics: Outbreak Modeling, Branching Process, Model Validation, Reproducibility)
- [What Is the Role of AI in Peer Review?](https://ghrbook.com/notes/ai-peer-review.html): A Small Illustration Using Open Peer Review. A blind panel of AI reviewers and the real human referees of a published BMJ trial reached the same verdict and overlapped on the essentials, but were strongest in different places—the AI on mechanical and citation checks, the humans on judgment. A worked example of what an AI reviewer adds to, and misses from, peer review. (Ch. 21 Publishing; topics: Peer Review, AI in Research, Open Science, Research Integrity) (Research Notes podcast episode with Tim Feeney and Navjoyt Ladher, The BMJ)
- [Shifting the True Endpoint](https://ghrbook.com/notes/surrogate-endpoints-survival.html): Major Progress in Pancreatic Cancer Treatment. In the 2026 RASolute 302 trial, daraxonrasib improved overall survival in metastatic pancreatic cancer, a shift in the true endpoint rather than a surrogate. Use it to separate surrogate endpoints from survival, see why most oncology approvals lean on surrogates, and learn what a randomized survival comparison can and cannot tell you. (Ch. 9 Measurement; topics: Surrogate Endpoints, Overall Survival, Randomized Trials, Outcome Measurement)
- [Testing Ebola Drugs in an Epidemic](https://ghrbook.com/notes/ebola-trial-methods-therapeutics.html): How a decade of Ebola research produced licensed treatments — and what happens next. Trace how Ebola therapeutic trials evolved from single-arm triage designs to a four-arm adaptive platform trial run in an active conflict zone, and what the resulting WHO recommendations can and cannot tell us about the 2026 Bundibugyo outbreak. (Ch. 10 Experimental Designs; topics: Adaptive Trials, Platform Trials, Single-Arm Designs, Outbreak Research)
- [Testing Ebola Vaccines in an Epidemic](https://ghrbook.com/notes/ebola-trial-methods-vaccines.html): How the same crisis produced two licensed vaccines — by two very different standards of evidence. Ervebo's efficacy was measured in humans through a ring-vaccination trial; Janssen's two-dose regimen was inferred from animal challenge via immunobridging. Follow how that distinction was made — and what each design choice cost in interpretability. (Ch. 10 Experimental Designs; topics: Ring Vaccination, Cluster-Randomized Trials, Immunobridging, Outbreak Research) (Research Notes podcast episode with Dr. Ana Maria Henao-Restrepo, World Health Organization (former))
- [When Should a Trial Stop?](https://ghrbook.com/notes/adaptive-trial.html): Bayesian Adaptive Design in a Kenyan Eye Care Program. How a research team embedded a Bayesian adaptive trial directly inside a mobile eye-screening app in Kenya, why it stopped after just four weeks, and what the design tradeoff between speed and certainty looks like in real-world implementation research. (Ch. 10 Experimental Designs; topics: Adaptive Trials, Bayesian Methods, Implementation Science, Randomized Trials) (Research Notes podcast episode with Dr. David Macleod)
- [Is ChatGPT a Substitute for Research Participants?](https://ghrbook.com/notes/synthetic-control.html): Making Sense of Silicon Subjects, Digital Twins, and Synthetic Control. Separate three concepts being conflated in public discourse — synthetic participants, the synthetic control method, and AI-augmented inference — using South Africa's tobacco tax policy as a worked example of how synthetic control builds counterfactuals from real data. (Ch. 11 Quasi-Experimental Designs; topics: Synthetic Control, Causal Inference, Digital Twins, AI in Research) (Research Notes podcast episode with Dr. Grieve Chelwa)
- [When an Algorithm Encodes Inequity](https://ghrbook.com/notes/race-kidney-transplants.html): Using Interrupted Time Series to Evaluate a National Policy to Undo Race-Based Harm in Kidney Transplantation. Use a recent study of the OPTN wait time modification policy to learn how interrupted time series designs evaluate policies implemented at a known point in time, with a comparison group, washout period, and real OPTN data. (Ch. 11 Quasi-Experimental Designs; topics: Interrupted Time Series, Health Equity, Kidney Transplantation, Policy Evaluation) (Research Notes podcast episode with Dr. Rohan Khazanchi)
- [Can You Blind a Psychedelic?](https://ghrbook.com/notes/blinding-psychedelics.html): Functional Unblinding, Expectancy Bias, and What It Means for Psychedelic Trial Design. Use Compass Pathways' psilocybin trials to explore how functional unblinding challenges the interpretation of psychedelic trial results, and what sensitivity analysis, interaction models, and design innovations can tell us about the boundary between pharmacology and expectancy. (Ch. 10 Experimental Designs; topics: Blinding, Expectancy Bias, Psychedelic Trials, Sensitivity Analysis) (Research Notes podcast episode with Dr. Gabe Loewinger, National Institute of Mental Health)
- [What Does 'Clinically Meaningful' Mean?](https://ghrbook.com/notes/clinically-meaningful.html): MADRS Thresholds, MCID, and the Difference Between Statistical and Clinical Significance. Use Compass Pathways' Phase 3 psilocybin trials to examine how remission thresholds, response definitions, and the minimum clinically important difference shape the interpretation of trial results. (Ch. 6 Statistical Inference; Ch. 10 Experimental Designs; topics: Clinical Significance, MCID, Depression Trials, Outcome Measurement) (Research Notes podcast episode with Dr. Jerry Rosenbaum, MGH Center for the Neuroscience of Psychedelics)
- [What Counts as a 'Clinical Trial'?](https://ghrbook.com/notes/clinical-trial-definition.html): How the NIH's Shifting Definition Reshapes Transparency in Research. Use the NIH's 2026 decision to exclude basic experimental studies from its clinical trial definition to explore what counts as a trial, why definitions matter for transparency, and how ClinicalTrials.gov registration has shaped modern research. (Ch. 10 Experimental Designs; topics: Trial Registration, ClinicalTrials.gov, NIH Policy, Research Transparency)
- [What Makes a Trial 'Adequate and Well-Controlled'?](https://ghrbook.com/notes/adequate-and-well-controlled.html): Lessons from the FDA's Refusal to Review an mRNA Flu Vaccine. Use the FDA's 2026 refusal to review Moderna's mRNA flu vaccine to understand why comparator choice defines what a trial can prove — and how regulatory, clinical, and global health considerations shape the meaning of 'adequate and well-controlled.' (Ch. 10 Experimental Designs; topics: Trial Design, Active Controls, FDA Regulation, Comparator Choice)
- [Drawing Your Assumptions](https://ghrbook.com/notes/dags-causal-inference.html): How a DAG Guided a Real Causal Analysis. Walk through a real research example to see how directed acyclic graphs (DAGs) help you identify confounders, select adjustment sets, and justify your analytical strategy. (Ch. 7 Causal Inference; topics: DAGs, Confounding, Adjustment Sets, Causal Analysis) (Research Notes podcast episode with Dr. Judith Lieber, London School of Hygiene and Tropical Medicine)
- [Non-Significant ≠ No Effect](https://ghrbook.com/notes/significance-vs-effect.html): A Permutation-Based Demonstration. Use a permutation-based simulation to understand why a non-significant p-value does not mean there is no effect — and why effect sizes and confidence intervals tell a richer story. (Ch. 6 Statistical Inference; topics: P-values, Effect Sizes, Permutation Tests, Statistical Power)

## Chapter videos

> Short async lecture videos, one chapter at a time. Every video publishes its slides as HTML (not images) and a full transcript, so the content is readable without watching. 48 videos are currently published.

### [Chapter 1: Global Health Research](https://ghrbook.com/videos/chapters/1/)

- [What global health means, and why equity is in the definition](https://ghrbook.com/videos/what-global-health-means/): video 1, 7 min. Where the term "global health" came from, how it differs from international health and public health, and why every serious definition carries equity inside it rather than alongside it. Watch on YouTube (https://www.youtube.com/watch?v=Fmf87l0G_74). Slides: https://ghrbook.com/videos/what-global-health-means/slides.html. Transcript: https://ghrbook.com/videos/what-global-health-means/transcript.txt. (topics: Global Health, Health Equity, Definitions)
- [What research is, and what makes it scientific](https://ghrbook.com/videos/what-makes-research-scientific/): video 2, 7 min. Research is a systematic investigation designed to produce generalizable knowledge. Four features make it scientific: the approach is empirical, the procedures are public, the goal is inference, and the conclusions are uncertain. Watch on YouTube (https://www.youtube.com/watch?v=LzQqn-ERk0Q). Slides: https://ghrbook.com/videos/what-makes-research-scientific/slides.html. Transcript: https://ghrbook.com/videos/what-makes-research-scientific/transcript.txt. (topics: Scientific Method, Inference, Uncertainty, Research Design)
- [The research landscape: basic, applied, and the clinical trial pipeline](https://ghrbook.com/videos/basic-applied-and-clinical-research/): video 3, 9 min. How basic, applied, and clinical research divide the work of global health, walked through the phases of a clinical trial and the three-decade development of the first malaria vaccine. Watch on YouTube (https://www.youtube.com/watch?v=WDTrrVp41RI). Slides: https://ghrbook.com/videos/basic-applied-and-clinical-research/slides.html. Transcript: https://ghrbook.com/videos/basic-applied-and-clinical-research/transcript.txt. (topics: Basic Research, Applied Research, Clinical Trials, Malaria Vaccine)
- [From efficacy to impact: translation, implementation, and evaluation](https://ghrbook.com/videos/from-efficacy-to-impact/): video 4, 8 min. What happens after a trial shows that something works: the T1 to T4 translational stages, implementation and policy research, and the difference between monitoring a program and evaluating it. Watch on YouTube (https://www.youtube.com/watch?v=3_bEI_a1R8U). Slides: https://ghrbook.com/videos/from-efficacy-to-impact/slides.html. Transcript: https://ghrbook.com/videos/from-efficacy-to-impact/transcript.txt. (topics: Translational Research, Implementation Science, Monitoring and Evaluation, Policy Research)
- [Who pays, who decides, and where the work lives](https://ghrbook.com/videos/who-pays-who-decides/): video 5, 5 min. Who funds global health research, who sets its agenda, and where the findings get published — and what it means for the field that the money, the agenda, and the access to results all concentrate in the same places. Watch on YouTube (https://www.youtube.com/watch?v=IjNj7db2Y4o). Slides: https://ghrbook.com/videos/who-pays-who-decides/slides.html. Transcript: https://ghrbook.com/videos/who-pays-who-decides/transcript.txt. (topics: Research Funding, Research Agenda, Open Access, Equity)

### [Chapter 2: Build Collaborations](https://ghrbook.com/videos/chapters/2/)

- [Why partnership is a design problem](https://ghrbook.com/videos/partnership-as-a-design-problem/): video 1, 5 min. Global health research is done in teams, and those teams are often unequal before anyone sits down at the table. Where that inequality comes from, what the calls to decolonize global health are asking for, and what team science says about designing a collaboration rather than assuming one. Watch on YouTube (https://www.youtube.com/watch?v=wbQNGf1EoyU). Slides: https://ghrbook.com/videos/partnership-as-a-design-problem/slides.html. Transcript: https://ghrbook.com/videos/partnership-as-a-design-problem/transcript.txt. (topics: Collaboration, Decolonizing Global Health, Team Science, Community-Based Participatory Research)
- [Whose vision gets funded](https://ghrbook.com/videos/whose-vision-gets-funded/): video 2, 5 min. A team needs a shared vision, but a vision recruited one person at a time and a vision forged by a crisis produce very different collaborations. How teams develop through forming, storming, norming, and performing, and why funding decides whose vision becomes reality. Watch on YouTube (https://www.youtube.com/watch?v=ekcGA_caIPg). Slides: https://ghrbook.com/videos/whose-vision-gets-funded/slides.html. Transcript: https://ghrbook.com/videos/whose-vision-gets-funded/transcript.txt. (topics: Shared Vision, Team Development, Research Funding, Trust)
- [Conflict, leadership, and credit](https://ghrbook.com/videos/conflict-leadership-and-credit/): video 3, 6 min. Effective research teams tend to have more conflict, not less, because the work invites disagreement. Keeping scientific conflict from turning personal, what leadership and mentoring look like when any team member can lead, and the problem of crediting collective work through individual prizes and author lists. Watch on YouTube (https://www.youtube.com/watch?v=4znIZ7rKaRU). Slides: https://ghrbook.com/videos/conflict-leadership-and-credit/slides.html. Transcript: https://ghrbook.com/videos/conflict-leadership-and-credit/transcript.txt. (topics: Conflict, Leadership, Mentoring, Authorship)
- [The Partnership Pathway: how it nearly fell apart](https://ghrbook.com/videos/the-partnership-pathway/): video 4, 6 min. In the 1990s, multidrug-resistant tuberculosis was widely regarded as untreatable at scale in low-income countries. The partnership that set out to prove otherwise: how it formed through a personal network rather than a call for proposals, how it was structured, and where it started to come apart. Watch on YouTube (https://www.youtube.com/watch?v=meLRPxVhNqA). Slides: https://ghrbook.com/videos/the-partnership-pathway/slides.html. Transcript: https://ghrbook.com/videos/the-partnership-pathway/transcript.txt. (topics: Partnership Pathway, MDR-TB, Case Study, Partners In Health)
- [What held it together, and the hard look in the mirror](https://ghrbook.com/videos/what-held-the-partnership-together/): video 5, 5 min. The MDR-TB partnership worked, and its treatment approach was integrated into global policy. What actually held it together, and the harder question underneath: good principles can still operate inside an inequitable structure, so it is worth asking who held the authority throughout. Watch on YouTube (https://www.youtube.com/watch?v=svxJr4V0dFk). Slides: https://ghrbook.com/videos/what-held-the-partnership-together/slides.html. Transcript: https://ghrbook.com/videos/what-held-the-partnership-together/transcript.txt. (topics: Trust, Equity, Research Partnerships, MDR-TB)

### [Chapter 3: Develop Research Ideas and Questions](https://ghrbook.com/videos/chapters/3/)

- [Where good ideas come from](https://ghrbook.com/videos/where-good-ideas-come-from/): video 1, 4 min. Good ideas are rarely eureka moments. The slow hunch and the adjacent possible explain why a research idea takes longer than a semester allows, and why opening one door tends to reveal several more. Watch on YouTube (https://www.youtube.com/watch?v=0-4d8S_gABs). Slides: https://ghrbook.com/videos/where-good-ideas-come-from/slides.html. Transcript: https://ghrbook.com/videos/where-good-ideas-come-from/transcript.txt. (topics: Research Ideas, Slow Hunch, Adjacent Possible, Creativity)
- [Finding a research problem worth studying](https://ghrbook.com/videos/finding-a-research-problem/): video 2, 5 min. What separates a research problem from a topic: it has to be solvable by systematic public methods. Bed nets prevent malaria, but we do not know how to get families to use them every night — and that gap is where a study lives. Watch on YouTube (https://www.youtube.com/watch?v=RBXp2Fs5vx0). Slides: https://ghrbook.com/videos/finding-a-research-problem/slides.html. Transcript: https://ghrbook.com/videos/finding-a-research-problem/transcript.txt. (topics: Research Problems, Literature Review, Problem Selection)
- [Description as a research goal](https://ghrbook.com/videos/description-as-a-research-goal/): video 3, 4 min. Descriptive statistics are not descriptive research. What it means to characterize a distribution in a well-defined population, worked through Nepal’s 2011 survey data on modern contraceptive use. Watch on YouTube (https://www.youtube.com/watch?v=prX1wIkzCKk). Slides: https://ghrbook.com/videos/description-as-a-research-goal/slides.html. Transcript: https://ghrbook.com/videos/description-as-a-research-goal/transcript.txt. (topics: Descriptive Research, Learning Goals, Survey Data, Epidemiology)
- [Explain and predict: why, what if, what comes next](https://ghrbook.com/videos/explanatory-and-predictive-questions/): video 4, 7 min. The two questions description cannot answer. Qualitative inquiry for why, causal inference for what if, and prediction for what comes next — with a warning about what "predictor" means in a regression table. Watch on YouTube (https://www.youtube.com/watch?v=ugyVCu2zDOM). Slides: https://ghrbook.com/videos/explanatory-and-predictive-questions/slides.html. Transcript: https://ghrbook.com/videos/explanatory-and-predictive-questions/transcript.txt. (topics: Explanation, Causal Inference, Qualitative Research, Prediction)
- [From question to hypothesis](https://ghrbook.com/videos/from-question-to-hypothesis/): video 5, 8 min. Answerable does not mean easy. FINER and PICO as checklists for specifying a question, why not all research needs a hypothesis, and why the hypotheses that matter are falsifiable and come from a theory that says why. Watch on YouTube (https://www.youtube.com/watch?v=L9lirnOBKVs). Slides: https://ghrbook.com/videos/from-question-to-hypothesis/slides.html. Transcript: https://ghrbook.com/videos/from-question-to-hypothesis/transcript.txt. (topics: Research Questions, FINER, PICO, Hypotheses, Theory)

### [Chapter 4: Searching the Literature](https://ghrbook.com/videos/chapters/4/)

- [What kind of review are you doing?](https://ghrbook.com/videos/what-kind-of-review/): video 1, 5 min. PubMed alone indexes 36 million citations, and that is only the indexed English-language part. Systematic, scoping, narrative, and bibliometric reviews as four different questions you can ask of a literature. Watch on YouTube (https://www.youtube.com/watch?v=flwqfbEd2Rw). Slides: https://ghrbook.com/videos/what-kind-of-review/slides.html. Transcript: https://ghrbook.com/videos/what-kind-of-review/transcript.txt. (topics: Literature Review, Systematic Reviews, Scoping Reviews, Bibliometrics)
- [Start with what someone already synthesized](https://ghrbook.com/videos/start-with-a-synthesis/): video 2, 5 min. Cochrane, Campbell, and WHO guidelines are places the work has already been done. The evidence hierarchy, and the qualification that matters: the best evidence depends on the question, and a poorly conducted trial does not beat a rigorous case-control study. Watch on YouTube (https://www.youtube.com/watch?v=iAHpTmQeO9s). Slides: https://ghrbook.com/videos/start-with-a-synthesis/slides.html. Transcript: https://ghrbook.com/videos/start-with-a-synthesis/transcript.txt. (topics: Evidence Synthesis, Cochrane, Evidence Hierarchy, Google Scholar)
- [Emerging AI tools, and being wary of fake science](https://ghrbook.com/videos/ai-tools-and-fake-science/): video 3, 5 min. AI discovery tools and predatory journals are the same failure wearing different clothes: something that looks like a paper and is not. What Elicit, Consensus, and Scite are good for, what hallucinated citations cost, and how to spot a predatory journal. Watch on YouTube (https://www.youtube.com/watch?v=A30Qn2qOkkE). Slides: https://ghrbook.com/videos/ai-tools-and-fake-science/slides.html. Transcript: https://ghrbook.com/videos/ai-tools-and-fake-science/transcript.txt. (topics: AI Tools, Predatory Journals, Research Integrity, Search Strategy)
- [How to conduct a systematic review, step by step](https://ghrbook.com/videos/conduct-a-systematic-review/): video 4, 13 min. The pivot from enough to everything, because the studies you miss are systematically different from the ones you find easily. Step 0 through Step 10 in one pass: a librarian as the first hire, PICO, controlled vocabulary, Boolean logic, screening, appraisal, and the PRISMA flow diagram. Watch on YouTube (https://www.youtube.com/watch?v=cb0GqhwYfyg). Slides: https://ghrbook.com/videos/conduct-a-systematic-review/slides.html. Transcript: https://ghrbook.com/videos/conduct-a-systematic-review/transcript.txt. (topics: Systematic Review, PICO, Boolean Search, PRISMA, Screening)
- [A Cochrane review, start to finish](https://ghrbook.com/videos/a-cochrane-review-start-to-finish/): video 5, 8 min. One published review walked through as the ten steps actually executed: drugs for preventing malaria in pregnancy, from 181 records down to 17 trials covering 14,481 women across eight countries and five decades. Watch on YouTube (https://www.youtube.com/watch?v=UM6mKK8awHU). Slides: https://ghrbook.com/videos/a-cochrane-review-start-to-finish/slides.html. Transcript: https://ghrbook.com/videos/a-cochrane-review-start-to-finish/transcript.txt. (topics: Cochrane Review, Worked Example, Malaria, Meta-Analysis)

### [Chapter 5: How to Read Scientific Articles](https://ghrbook.com/videos/chapters/5/)

- [What kind of paper is this?](https://ghrbook.com/videos/what-kind-of-paper-is-this/): video 1, 6 min. Reading a paper and appraising one are different skills. Where evidence-based medicine came from, and the first skill it asks for: classifying a study by design in about a minute. Watch on YouTube (https://www.youtube.com/watch?v=YLIfh9hgIk8). Slides: https://ghrbook.com/videos/what-kind-of-paper-is-this/slides.html. Transcript: https://ghrbook.com/videos/what-kind-of-paper-is-this/transcript.txt. (topics: Critical Appraisal, Study Design, Evidence-Based Medicine)
- [Read the methods first, and find where bias enters](https://ghrbook.com/videos/read-the-methods-first/): video 2, 6 min. Question, design, participants, comparison, outcomes, transparency — the frame for appraising any study. Then the hard idea: confounding is what you failed to adjust for, selection bias is who ended up in the study, and only one of them can be fixed by statistics. Watch on YouTube (https://www.youtube.com/watch?v=MGP7AjJOJgQ). Slides: https://ghrbook.com/videos/read-the-methods-first/slides.html. Transcript: https://ghrbook.com/videos/read-the-methods-first/transcript.txt. (topics: Methods, Confounding, Selection Bias, Critical Appraisal)
- [A real methods section, read out loud](https://ghrbook.com/videos/a-methods-section-read-out-loud/): video 3, 7 min. A published methods section worked through line by line, with the reader’s own questions left in: why randomize districts rather than children, what a two-thirds completion rate does to the comparison, and why particular outcomes were named in advance. Watch on YouTube (https://www.youtube.com/watch?v=NSKAQT9YDtQ). Slides: https://ghrbook.com/videos/a-methods-section-read-out-loud/slides.html. Transcript: https://ghrbook.com/videos/a-methods-section-read-out-loud/transcript.txt. (topics: Methods, Malaria Vaccine, Worked Example, Study Design)
- [Effect sizes, confidence intervals, and how results get misread](https://ghrbook.com/videos/effect-sizes-and-confidence-intervals/): video 4, 5 min. Why the interval carries more information than the p-value, worked through two risk ratios where a non-significant result is consistent with a 40% reduction, no effect, and a 5% increase at the same time. Watch on YouTube (https://www.youtube.com/watch?v=aNsfpjwnNNk). Slides: https://ghrbook.com/videos/effect-sizes-and-confidence-intervals/slides.html. Transcript: https://ghrbook.com/videos/effect-sizes-and-confidence-intervals/transcript.txt. (topics: Effect Size, Confidence Intervals, p-values, Interpretation)
- [Reading tables and figures](https://ghrbook.com/videos/reading-tables-and-figures/): video 5, 9 min. Four real figures from one study: a baseline table with shifting denominators, a regression table with unadjusted against adjusted odds ratios, Kaplan-Meier curves, and a flow diagram that starts at 12,064 patients and ends at 10,102. Watch on YouTube (https://www.youtube.com/watch?v=BN0pEG2L15U). Slides: https://ghrbook.com/videos/reading-tables-and-figures/slides.html. Transcript: https://ghrbook.com/videos/reading-tables-and-figures/transcript.txt. (topics: Tables, Figures, Regression, Kaplan-Meier, Forest Plots)
- [Qualitative rigor, and reading the discussion like a skeptic](https://ghrbook.com/videos/qualitative-rigor-and-the-discussion/): video 6, 9 min. Qualitative work judged on its own terms — informational richness rather than representativeness, and why one quote from one person is an anecdote rather than a theme. Then the discussion section: the slip from associated with to leads to, and what authors leave out. Watch on YouTube (https://www.youtube.com/watch?v=zsHrRUF9tnI). Slides: https://ghrbook.com/videos/qualitative-rigor-and-the-discussion/slides.html. Transcript: https://ghrbook.com/videos/qualitative-rigor-and-the-discussion/transcript.txt. (topics: Qualitative Research, Saturation, Discussion Section, Limitations)

### [Chapter 6: Statistical Inference](https://ghrbook.com/videos/chapters/6/)

- [What is statistical inference?](https://ghrbook.com/videos/what-is-statistical-inference/): video 1, 7 min. A large trial found a 30% lower risk of cancer among older women taking vitamin D3 and calcium, and the authors still reported the result as inconclusive. Working through that disagreement introduces statistical conclusion validity and the two major approaches to inference. Watch on YouTube (https://www.youtube.com/watch?v=fk-yjmJIEww). Slides: https://ghrbook.com/videos/what-is-statistical-inference/slides.html. Transcript: https://ghrbook.com/videos/what-is-statistical-inference/transcript.txt. (topics: Statistical Inference, Statistical Conclusion Validity, Frequentist Statistics, Bayesian Statistics)
- [Imagine a world where the treatment does nothing](https://ghrbook.com/videos/imagine-the-treatment-does-nothing/): video 2, 9 min. Frequentist inference judges the one study you ran by comparing it against studies that were never run. Building that imaginary world one figure at a time — null hypothesis, simulation, and the Central Limit Theorem — with a trial of lay-counselor depression treatment in India as the example. Watch on YouTube (https://www.youtube.com/watch?v=-x2VMzwgz2E). Slides: https://ghrbook.com/videos/imagine-the-treatment-does-nothing/slides.html. Transcript: https://ghrbook.com/videos/imagine-the-treatment-does-nothing/transcript.txt. (topics: Null Hypothesis, Null Distribution, Central Limit Theorem, Simulation)
- [Retaining or rejecting the null hypothesis](https://ghrbook.com/videos/retaining-or-rejecting-the-null/): video 3, 6 min. Once the null distribution exists, the decision about your result is close to mechanical. Alpha as a long-run error rate chosen before the study is run, the p-value as a conditional probability, and false positives and false negatives shown across simulated studies. Watch on YouTube (https://www.youtube.com/watch?v=RMjHoZVz46U). Slides: https://ghrbook.com/videos/retaining-or-rejecting-the-null/slides.html. Transcript: https://ghrbook.com/videos/retaining-or-rejecting-the-null/transcript.txt. (topics: Significance Testing, p-values, Type I and Type II Error, Alpha Level)
- [What a p-value doesn’t tell you](https://ghrbook.com/videos/what-a-p-value-doesnt-tell-you/): video 4, 8 min. The p-value answers a narrower question than most readers think it does. Statistical against clinical significance, why published p-values pile up just under the threshold, and what effect sizes and confidence intervals offer in place of a yes-or-no verdict. Watch on YouTube (https://www.youtube.com/watch?v=KuCCl_ipwR4). Slides: https://ghrbook.com/videos/what-a-p-value-doesnt-tell-you/slides.html. Transcript: https://ghrbook.com/videos/what-a-p-value-doesnt-tell-you/transcript.txt. (topics: p-values, Effect Size, Confidence Intervals, Publication Bias)
- [The Bayesian approach](https://ghrbook.com/videos/the-bayesian-approach/): video 5, 8 min. A positive rapid COVID test means something different early in a pandemic than it does during a surge, and the difference is the base rate you started from. Bayesian inference built from that intuition, then the same trial analyzed both ways — nearly identical numbers, different interpretations. Watch on YouTube (https://www.youtube.com/watch?v=veOufiHR01A). Slides: https://ghrbook.com/videos/the-bayesian-approach/slides.html. Transcript: https://ghrbook.com/videos/the-bayesian-approach/transcript.txt. (topics: Bayesian Statistics, Priors and Posteriors, Credible Intervals, Base Rates)

### [Chapter 7: Causal Inference](https://ghrbook.com/videos/chapters/7/)

- [Causal deniability](https://ghrbook.com/videos/causal-deniability/): video 1, 8 min. A large study of coffee drinkers reported a 5% lower risk of dying, stated plainly that it could not establish cause and effect, and then suggested people drink coffee for their health. Naming that move, and the three conditions for demonstrating causality. Watch on YouTube (https://www.youtube.com/watch?v=tghfv9_3S_s). Slides: https://ghrbook.com/videos/causal-deniability/slides.html. Transcript: https://ghrbook.com/videos/causal-deniability/transcript.txt. (topics: Causal Inference, Association and Causation, Confounding, Internal Validity)
- [The road not taken](https://ghrbook.com/videos/the-road-not-taken/): video 2, 9 min. A causal effect is a difference between two outcomes for the same person, and only one of them ever happens. Potential outcomes and the fundamental problem of causal inference, worked through a population whose true average treatment effect is known to be 1.5. Watch on YouTube (https://www.youtube.com/watch?v=TkL_XvVUrak). Slides: https://ghrbook.com/videos/the-road-not-taken/slides.html. Transcript: https://ghrbook.com/videos/the-road-not-taken/transcript.txt. (topics: Potential Outcomes, Counterfactuals, Average Treatment Effect, Selection Bias)
- [Confounder control and drawing your assumptions](https://ghrbook.com/videos/confounder-control-and-causal-diagrams/): video 3, 7 min. Ice cream sales and violent crime rise together, and no amount of staring at the data will settle why — a dataset carries no memory of what produced it. A causal diagram is where you write the story down before estimation starts, built on forks, pipes, and colliders. Watch on YouTube (https://www.youtube.com/watch?v=5q8Rc3skZOs). Slides: https://ghrbook.com/videos/confounder-control-and-causal-diagrams/slides.html. Transcript: https://ghrbook.com/videos/confounder-control-and-causal-diagrams/transcript.txt. (topics: Causal Diagrams, DAGs, Confounding, Collider Bias)
- [Good paths and bad paths](https://ghrbook.com/videos/good-paths-and-bad-paths/): video 4, 5 min. Between any two variables in a causal diagram there are usually several routes, and identification is the discipline of knowing which ones carry the causal effect and which carry confounding. Path tracing, minimum sufficient adjustment sets, d-separation, and bad controls. Watch on YouTube (https://www.youtube.com/watch?v=n8TFDTLlsVA). Slides: https://ghrbook.com/videos/good-paths-and-bad-paths/slides.html. Transcript: https://ghrbook.com/videos/good-paths-and-bad-paths/transcript.txt. (topics: Backdoor Paths, Effect Identification, d-separation, Adjustment Sets)
- [Closing backdoor paths](https://ghrbook.com/videos/closing-backdoor-paths/): video 5, 9 min. A simulated dataset where the true effect is 1.5 by construction: the naive comparison returns 2.07, holding the confounder constant returns 1.47, and the specification that adds a collider moves the estimate away from the truth. More control, worse answer. Watch on YouTube (https://www.youtube.com/watch?v=c13aEjwTdrw). Slides: https://ghrbook.com/videos/closing-backdoor-paths/slides.html. Transcript: https://ghrbook.com/videos/closing-backdoor-paths/transcript.txt. (topics: Regression Adjustment, Confounding, Collider Bias, Causal Assumptions)
- [Instrument-based approaches](https://ghrbook.com/videos/instrument-based-approaches/): video 6, 9 min. Randomization destroys confounding by construction, because nothing upstream can influence an assignment made by a coin. When a coin is not available, the alternative is variation in the world that behaves like one — here, an HPV vaccination program where eligibility turned on a birth date. Watch on YouTube (https://www.youtube.com/watch?v=-8ZQjFs2wGs). Slides: https://ghrbook.com/videos/instrument-based-approaches/slides.html. Transcript: https://ghrbook.com/videos/instrument-based-approaches/transcript.txt. (topics: Instrumental Variables, Regression Discontinuity, Natural Experiments, HPV Vaccination)

### [Chapter 8: External Validity, Generalizability, and Transportability](https://ghrbook.com/videos/chapters/8/)

- [External validity, generalizability, and transportability](https://ghrbook.com/videos/external-validity-generalizability-transportability/): video 1, 5 min. Clinical Trial 320 showed that three-drug combination therapy cut the risk of AIDS or death roughly in half; re-analyzed years later for a different population, the estimated benefit shrank. External validity as the umbrella, with generalizability and transportability underneath it. Watch on YouTube (https://www.youtube.com/watch?v=e_FYhqrD6HE). Slides: https://ghrbook.com/videos/external-validity-generalizability-transportability/slides.html. Transcript: https://ghrbook.com/videos/external-validity-generalizability-transportability/transcript.txt. (topics: External Validity, Generalizability, Transportability, Effect Modification)
- [Does the sample stand for the population it came from?](https://ghrbook.com/videos/does-the-sample-stand-for-the-population/): video 2, 10 min. Generalizability asks whether results from your sample stand for the population that sample came from, worked four times: an estimate carried by its sampling design, one carried by a model’s assumptions, findings honest about belonging to one village, and a causal effect whose reach depends on effect modification. Watch on YouTube (https://www.youtube.com/watch?v=zQXLJCdE36k). Slides: https://ghrbook.com/videos/does-the-sample-stand-for-the-population/slides.html. Transcript: https://ghrbook.com/videos/does-the-sample-stand-for-the-population/transcript.txt. (topics: Generalizability, Probability Sampling, Poststratification, Survey Design)
- [Do the results apply to a different population entirely?](https://ghrbook.com/videos/do-results-apply-to-another-population/): video 3, 10 min. Transportability asks whether results apply to a population your sample was never part of, and no amount of better sampling fixes it. A 1997 HIV trial carried to a later population by re-expressing its age-specific effects against that population’s composition. Watch on YouTube (https://www.youtube.com/watch?v=uHNJI7GVXNo). Slides: https://ghrbook.com/videos/do-results-apply-to-another-population/slides.html. Transcript: https://ghrbook.com/videos/do-results-apply-to-another-population/transcript.txt. (topics: Transportability, Standardization, Effect Modification, Hazard Ratios)
- [Designing for external validity](https://ghrbook.com/videos/designing-for-external-validity/): video 4, 10 min. External validity as a design decision made before enrollment rather than an analysis run afterward, then Shadish, Cook and Campbell’s five threats as a checklist you can put to your own design or to someone else’s paper. Watch on YouTube (https://www.youtube.com/watch?v=X09LTfvGYnA). Slides: https://ghrbook.com/videos/designing-for-external-validity/slides.html. Transcript: https://ghrbook.com/videos/designing-for-external-validity/transcript.txt. (topics: External Validity, Study Design, Threats to Validity, Efficacy and Effectiveness)

### [Chapter 9: Measurement and Construct Validation](https://ghrbook.com/videos/chapters/9/)

- [Even counting the dead takes judgment](https://ghrbook.com/videos/even-counting-the-dead-takes-judgment/): video 1, 6 min. Deciding whether someone died with COVID-19 or from it is a judgment made by a certifier, and the proposed fix — excess mortality — brings challenges of its own. Construct, indicator, instrument, and measure: the vocabulary the rest of the chapter runs on. Watch on YouTube (https://www.youtube.com/watch?v=PNQIN3TfTNw). Slides: https://ghrbook.com/videos/even-counting-the-dead-takes-judgment/slides.html. Transcript: https://ghrbook.com/videos/even-counting-the-dead-takes-judgment/transcript.txt. (topics: Measurement, Construct Validity, Indicators, Excess Mortality)
- [From conceptual model to measurement](https://ghrbook.com/videos/from-conceptual-model-to-measurement/): video 2, 6 min. A conceptual model you drew for another purpose will tell you what to measure. A DAG turns a causal inference problem into a data collection checklist, and a logic model does the parallel job for a program — ending in treatment fidelity and the difference between implementation failure and theory failure. Watch on YouTube (https://www.youtube.com/watch?v=KFiyb3mOayU). Slides: https://ghrbook.com/videos/from-conceptual-model-to-measurement/slides.html. Transcript: https://ghrbook.com/videos/from-conceptual-model-to-measurement/transcript.txt. (topics: Conceptual Models, Logic Models, Treatment Fidelity, Monitoring and Evaluation)
- [Finding good indicators](https://ghrbook.com/videos/finding-good-indicators/): video 3, 5 min. DREAMY as a checklist for a good indicator, with two fully specified indicators of depression as the worked case. Includes what to do when the thing you care about cannot be put to a respondent as a question, such as measuring corruption by digging up a road. Watch on YouTube (https://www.youtube.com/watch?v=1PQO5ACvKOw). Slides: https://ghrbook.com/videos/finding-good-indicators/slides.html. Transcript: https://ghrbook.com/videos/finding-good-indicators/transcript.txt. (topics: Indicators, Operationalization, Composite Measures, Survey Design)
- [Indexes and scales](https://ghrbook.com/videos/indexes-and-scales/): video 4, 7 min. Indexes and scales both combine several items into one score, and they do it for opposite reasons. In an index the items cause the score; in a scale the score causes the items, and the direction of that arrow changes what you are allowed to do with the number. Watch on YouTube (https://www.youtube.com/watch?v=fLPTLMnZtic). Slides: https://ghrbook.com/videos/indexes-and-scales/slides.html. Transcript: https://ghrbook.com/videos/indexes-and-scales/transcript.txt. (topics: Indexes, Scales, Latent Variables, Psychometrics)
- [Validity is not a property of an instrument](https://ghrbook.com/videos/validity-is-not-a-property-of-an-instrument/): video 5, 8 min. A validated instrument is not a settled one. Validity is a judgment about whether scores mean what you think they mean — for these people, in this setting, for this purpose — worked through content validity, cognitive interviewing, and item analysis. Watch on YouTube (https://www.youtube.com/watch?v=6mAItKDYt34). Slides: https://ghrbook.com/videos/validity-is-not-a-property-of-an-instrument/slides.html. Transcript: https://ghrbook.com/videos/validity-is-not-a-property-of-an-instrument/transcript.txt. (topics: Construct Validation, Content Validity, Cognitive Interviewing, Item Analysis)
- [Factor analysis and reliability](https://ghrbook.com/videos/factor-analysis-and-reliability/): video 6, 10 min. Two questions about a set of items, asked in order. Factor analysis asks whether they measure one thing or several, and reliability asks whether they measure it consistently — test-retest, internal consistency, inter-rater agreement, and responsiveness to change. Watch on YouTube (https://www.youtube.com/watch?v=Ag1HaquXjqA). Slides: https://ghrbook.com/videos/factor-analysis-and-reliability/slides.html. Transcript: https://ghrbook.com/videos/factor-analysis-and-reliability/transcript.txt. (topics: Factor Analysis, Reliability, Internal Consistency, Psychometrics)
- [Does it match reality, and does it travel?](https://ghrbook.com/videos/does-it-match-reality-and-does-it-travel/): video 7, 6 min. The last phase of construct validation turns outward: do the scores line up with something in the world? Diagnostic accuracy read off a confusion matrix, then cross-cultural validity and the response formats that fail when an instrument crosses a border. Watch on YouTube (https://www.youtube.com/watch?v=N5Tc1sK8vUA). Slides: https://ghrbook.com/videos/does-it-match-reality-and-does-it-travel/slides.html. Transcript: https://ghrbook.com/videos/does-it-match-reality-and-does-it-travel/transcript.txt. (topics: Criterion Validity, Diagnostic Accuracy, Cross-Cultural Validity, Screening)

## Research Notes podcast

> A conversation series with the researchers behind the companion notes. Each episode pairs with a note above; listen, watch on YouTube, or read. Subscribe via Apple Podcasts, Spotify, YouTube, or RSS (https://anchor.fm/s/10f7865bc/podcast/rss).

- [What Would It Take to Stop This Outbreak?](https://ghrbook.com/notes/bundibugyo-outbreak-model.html): conversation with LCDR Eric Mooring, U.S. Centers for Disease Control and Prevention. Watch on YouTube (https://www.youtube.com/watch?v=61QGrkgH-yA) or read the companion note.
- [What Is the Role of AI in Peer Review?](https://ghrbook.com/notes/ai-peer-review.html): conversation with Tim Feeney and Navjoyt Ladher, The BMJ. Watch on YouTube (https://www.youtube.com/watch?v=aTGaucCAxww) or read the companion note.
- [Testing Ebola Vaccines in an Epidemic](https://ghrbook.com/notes/ebola-trial-methods-vaccines.html): conversation with Dr. Ana Maria Henao-Restrepo, World Health Organization (former). Watch on YouTube (https://www.youtube.com/watch?v=Zobs-8WBm0w) or read the companion note.
- [When Should a Trial Stop?](https://ghrbook.com/notes/adaptive-trial.html): conversation with Dr. David Macleod. Watch on YouTube (https://www.youtube.com/watch?v=Vlf70-JNGp8) or read the companion note.
- [Is ChatGPT a Substitute for Research Participants?](https://ghrbook.com/notes/synthetic-control.html): conversation with Dr. Grieve Chelwa. Watch on YouTube (https://www.youtube.com/watch?v=mOAe9GqUO6w) or read the companion note.
- [When an Algorithm Encodes Inequity](https://ghrbook.com/notes/race-kidney-transplants.html): conversation with Dr. Rohan Khazanchi. Watch on YouTube (https://www.youtube.com/watch?v=e3DQYsoxN8U) or read the companion note.
- [Can You Blind a Psychedelic?](https://ghrbook.com/notes/blinding-psychedelics.html): conversation with Dr. Gabe Loewinger, National Institute of Mental Health. Watch on YouTube (https://www.youtube.com/watch?v=fs6zz3Gzv-w) or read the companion note.
- [What Does 'Clinically Meaningful' Mean?](https://ghrbook.com/notes/clinically-meaningful.html): conversation with Dr. Jerry Rosenbaum, MGH Center for the Neuroscience of Psychedelics. Watch on YouTube (https://www.youtube.com/watch?v=Eb3dQgJ2Coo) or read the companion note.
- [Drawing Your Assumptions](https://ghrbook.com/notes/dags-causal-inference.html): conversation with Dr. Judith Lieber, London School of Hygiene and Tropical Medicine. Watch on YouTube (https://www.youtube.com/watch?v=1Vg1qBroU6I) or read the companion note.