THE GOLD STANDARD OF INTELLIGENCE · SPECIALIST STUDY
Combine respondent-driven samples across sites only after deciding what population, design and estimand can legitimately be shared
Multi-site respondent-driven sampling does not become one larger representative sample merely because every site used coupons and the same questionnaire. Sites may target different populations, define network size differently, recruit through different venues, experience distinct seed and homophily patterns, close at different times and support different estimands. The gold standard harmonises eligibility and variables without erasing local meaning, preserves site-specific recruitment diagnostics and weights, states whether the target is a set of site estimates or a pooled population, tests transportability and effect heterogeneity, uses hierarchical models only under explicit exchangeability assumptions and retains a correction route when one site cannot support the shared claim.
This guide belongs to the Intelligence subject guide, the Gold Standard of Education, and the How X Works learning library. Choose a reading route, work the examples, then use the transfer checks to test whether the method survives a changed context.
Choose a reading route
Build the foundations · chapters 1–2
Apply the core method · chapters 3–5
Test evidence and alternatives · chapters 6–8
Handle limits and disagreement · chapters 9–11
Transfer and govern · chapters 12–13
The whole problem at a glance
| Question | Site estimate | Pooled estimate | Required evidence |
|---|---|---|---|
| Local prevalence | Primary | Not needed | Site design diagnostics |
| Average across named sites | Input | Defined weighting | Target-population weights |
| Common association | Stratified | Model-based | Heterogeneity checks |
| New-site prediction | Validation target | Hierarchical | Transportability evidence |
More participants do not repair a mismatch in target population.
CHAPTER 1 OF 13 · Build the foundations
1. Define the site populations and shared target
Core idea. Each RDS site samples through a local social network and fieldwork system. A multi-site analysis must state whether the target is each local population, an average of named sites, a larger super-population or a relationship expected to generalise.
Method. Write eligibility, geography, time, network boundary, recruitment setting and target estimand for every site; then identify the population measure that gives each site its contribution to a combined claim.
Worked example. Four city samples estimate four city prevalences. A national percentage is not identified unless the cities and their weights represent the national target.
Changed-context transfer. Changed-context transfer: apply the same reasoning structure to a school or public-service decision. Preserve the formal relationship, then replace the population, authority, privacy duties, evidence and remedy; never copy the domain conclusion.
Limits. Hidden populations may have no census denominators, site catchments can overlap and mobility can blur network boundaries.
Evidence and review. For this chapter, keep a dated evidence record: the input, rule or assumption, result, accountable owner and a condition that would change the conclusion. Work an ordinary case and a boundary case, preserve rejected alternatives, and require an authorised independent reader to reproduce where judgement entered. Chapter 1 is complete only when the claimed result survives that review.
Misconception to avoid. Pooling site samples automatically estimates the population across all sites. Sample size weights answer only one arbitrary average.
Practical evidence check
- Ask which person in the target population each pooled percentage represents and how their site weight was obtained.
- Return to chapter contents
CHAPTER 2 OF 13 · Build the foundations
2. Harmonise constructs without flattening local meaning
Core idea. Identical variable names can conceal different wording, recall windows, languages, eligibility screens or interviewer practices. Harmonisation needs item-level equivalence and explicit non-comparability codes.
Method. Build a crosswalk of source items, translations, response options, derivations, missingness and effective dates; retain site-local variables and create a shared variable only when the construct survives.
Worked example. Site A asks network size among people seen in 30 days; Site B asks people known in six months. Renaming both degree does not make inverse-degree weights comparable.
Changed-context transfer. Changed-context transfer: apply the same reasoning structure to a school or public-service decision. Preserve the formal relationship, then replace the population, authority, privacy duties, evidence and remedy; never copy the domain conclusion.
Limits. Perfect equivalence is rare and adaptation may be ethically necessary. Use sensitivity sets rather than forcing a false common variable.
Evidence and review. For this chapter, keep a dated evidence record: the input, rule or assumption, result, accountable owner and a condition that would change the conclusion. Work an ordinary case and a boundary case, preserve rejected alternatives, and require an authorised independent reader to reproduce where judgement entered. Chapter 2 is complete only when the claimed result survives that review.
Misconception to avoid. A common codebook proves harmonisation. Implementation and respondent interpretation can still differ.
Practical evidence check
- Blindly recode two site items, then have local teams explain which distinctions were lost.
- Return to chapter contents
CHAPTER 3 OF 13 · Apply the core method
3. Preserve site-specific RDS design information
Core idea. Seeds, coupons, recruitment trees, degree questions, incentive rules and closure decisions determine each site’s design evidence. Merging records before diagnostics destroys information needed to evaluate assumptions.
Method. Retain site, seed, recruiter, coupon issue and return, wave, recruitment time, degree version and fieldwork intervention; compute diagnostics separately before any shared model.
Worked example. One site has twenty seeds and shallow chains; another has four long chains. Their equal record format hides very different dependence and seed sensitivity.
Changed-context transfer. Changed-context transfer: apply the same reasoning structure to a school or public-service decision. Preserve the formal relationship, then replace the population, authority, privacy duties, evidence and remedy; never copy the domain conclusion.
Limits. Recruitment trees are partial networks, coupon failures are not all observed and cross-site recruitment can occur.
Evidence and review. For this chapter, keep a dated evidence record: the input, rule or assumption, result, accountable owner and a condition that would change the conclusion. Work an ordinary case and a boundary case, preserve rejected alternatives, and require an authorised independent reader to reproduce where judgement entered. Chapter 3 is complete only when the claimed result survives that review.
Misconception to avoid. The same number of participants means the same design quality. Effective information depends on network bottlenecks and recruitment dependence.
Practical evidence check
- Reconstruct one site’s recruitment tree from raw coupon events and verify no global merge altered parentage.
- Return to chapter contents
CHAPTER 4 OF 13 · Apply the core method
4. Separate local weights from cross-site weights
Core idea. RDS inclusion adjustments operate within a site’s network model; cross-site weights express how local target populations contribute to the combined estimand. Multiplying them without a defined target can create impressive but meaningless precision.
Method. Calculate local estimator weights under site-specific degree and sampling assumptions, define external site weights from the target population or policy estimand, normalise transparently and show every limiting case.
Worked example. A small city with a large sample should not dominate a population-weighted regional estimate merely because recruitment was efficient.
Changed-context transfer. Changed-context transfer: apply the same reasoning structure to a school or public-service decision. Preserve the formal relationship, then replace the population, authority, privacy duties, evidence and remedy; never copy the domain conclusion.
Limits. Reliable site population sizes may be unavailable, and local RDS weights themselves are model-dependent. Report alternative weighting schemes.
Evidence and review. For this chapter, keep a dated evidence record: the input, rule or assumption, result, accountable owner and a condition that would change the conclusion. Work an ordinary case and a boundary case, preserve rejected alternatives, and require an authorised independent reader to reproduce where judgement entered. Chapter 4 is complete only when the claimed result survives that review.
Misconception to avoid. Inverse-degree weights are enough for pooling. They address within-site selection, not the geography of the combined target.
Practical evidence check
- Compare participant-equal, site-equal and target-population-weighted estimates and explain the different questions.
- Return to chapter contents
CHAPTER 5 OF 13 · Apply the core method
5. Diagnose recruitment heterogeneity before modelling outcomes
Core idea. Sites can differ in homophily, recruitment effectiveness, seed composition, degree distribution and time shocks. Outcome differences may therefore mix population heterogeneity with design and fieldwork heterogeneity.
Method. Create a common diagnostic panel for chain depth, recruiter transition, coupon return, degree distribution, seed sensitivity, bottlenecks and calendar time; overlay local thresholds without forcing identical conclusions.
Worked example. Site B’s higher estimate disappears when two dominant seed chains are removed, while Site A remains stable across chains. Pooling first would hide the asymmetry.
Changed-context transfer. Changed-context transfer: apply the same reasoning structure to a school or public-service decision. Preserve the formal relationship, then replace the population, authority, privacy duties, evidence and remedy; never copy the domain conclusion.
Limits. Sensitivity analyses do not reveal the true unsampled network and small sites may have wide diagnostic uncertainty.
Evidence and review. For this chapter, keep a dated evidence record: the input, rule or assumption, result, accountable owner and a condition that would change the conclusion. Work an ordinary case and a boundary case, preserve rejected alternatives, and require an authorised independent reader to reproduce where judgement entered. Chapter 5 is complete only when the claimed result survives that review.
Misconception to avoid. A site indicator in regression controls all site differences. It does not repair selection or identify unmeasured design processes.
Practical evidence check
- Simulate a seed-dominated site and show how its pooled influence changes under declared sensitivity rules.
- Return to chapter contents
Illustrative site panel
| Site | Recruitment feature | Estimate | Interpretation |
|---|---|---|---|
| A | Long chains, moderate homophily | 18% | Design supported |
| B | Two dominant seeds | 31% | Seed sensitivity |
| C | Early closure | 12% | Chain censoring |
| D | Different degree wording | 24% | Not directly harmonised |
Numbers are synthetic and teach diagnosis, not a real population claim.
CHAPTER 6 OF 13 · Test evidence and alternatives
6. Choose a pooling model that matches the estimand
Core idea. Fixed site effects, random effects, meta-analysis, survey-weighted models and hierarchical Bayes answer different questions. The model must preserve site design, heterogeneity and the target weighting scheme.
Method. Specify observation model, local weights, site effects, variance estimator, dependence structure and target contrast; compare site-stratified, fixed-effect and partially pooled results.
Worked example. A hierarchical prevalence model shrinks a small site toward the group mean, while the published table still shows its unpooled estimate and explains the exchangeability assumption.
Changed-context transfer. Changed-context transfer: apply the same reasoning structure to a school or public-service decision. Preserve the formal relationship, then replace the population, authority, privacy duties, evidence and remedy; never copy the domain conclusion.
Limits. Standard errors from convenient software may ignore recruitment dependence or estimated weights. Small numbers of sites limit heterogeneity estimation.
Evidence and review. For this chapter, keep a dated evidence record: the input, rule or assumption, result, accountable owner and a condition that would change the conclusion. Work an ordinary case and a boundary case, preserve rejected alternatives, and require an authorised independent reader to reproduce where judgement entered. Chapter 6 is complete only when the claimed result survives that review.
Misconception to avoid. Random effects automatically make results generalisable. They formalise a distributional assumption about sites; they do not create a sampling frame of sites.
Practical evidence check
- Change the site hierarchy and prior or variance estimator, then identify which headline claims move.
- Return to chapter contents
CHAPTER 7 OF 13 · Test evidence and alternatives
7. Measure heterogeneity rather than averaging it away
Core idea. A pooled mean can be correct yet unhelpful when site effects are large or reverse direction. Heterogeneity is evidence about mechanisms, context and the limits of transfer.
Method. Report site estimates and uncertainty, predeclare effect modifiers, use interaction or meta-regression cautiously, and distinguish explained from residual heterogeneity.
Worked example. An intervention association is positive in two sites and negative in two. The near-zero pooled result does not mean no site has a consequential relationship.
Changed-context transfer. Changed-context transfer: apply the same reasoning structure to a school or public-service decision. Preserve the formal relationship, then replace the population, authority, privacy duties, evidence and remedy; never copy the domain conclusion.
Limits. With few sites, heterogeneity statistics are unstable and multiple subgroup searches invite selective stories.
Evidence and review. For this chapter, keep a dated evidence record: the input, rule or assumption, result, accountable owner and a condition that would change the conclusion. Work an ordinary case and a boundary case, preserve rejected alternatives, and require an authorised independent reader to reproduce where judgement entered. Chapter 7 is complete only when the claimed result survives that review.
Misconception to avoid. If the pooled interval is narrow, heterogeneity no longer matters. Precision around an average does not describe dispersion across sites.
Practical evidence check
- Draw a forest plot with the pooled target and prediction interval, then state what a new site could plausibly experience.
- Return to chapter contents
CHAPTER 8 OF 13 · Test evidence and alternatives
8. Account for dependence within chains and across sites
Core idea. Recruiter-recruit links create dependence within sites, while shared staff, venues, campaigns or mobile participants can create cross-site dependence. Treating all records as independent understates uncertainty.
Method. Cluster or model at defensible recruitment levels, retain seed and recruiter effects, identify shared operational shocks and use resampling units that preserve fieldwork structure.
Worked example. A regional campaign changes participation in three sites during the same month. Site-level calendar effects are correlated rather than independent replications.
Changed-context transfer. Changed-context transfer: apply the same reasoning structure to a school or public-service decision. Preserve the formal relationship, then replace the population, authority, privacy duties, evidence and remedy; never copy the domain conclusion.
Limits. The true social network is only partly observed, and few sites make robust cluster corrections difficult.
Evidence and review. For this chapter, keep a dated evidence record: the input, rule or assumption, result, accountable owner and a condition that would change the conclusion. Work an ordinary case and a boundary case, preserve rejected alternatives, and require an authorised independent reader to reproduce where judgement entered. Chapter 8 is complete only when the claimed result survives that review.
Misconception to avoid. A large pooled sample guarantees a small standard error. Dependence can leave the effective information far smaller than the row count.
Practical evidence check
- Compare naive, chain-clustered and site-shock-aware intervals for the same synthetic estimate.
- Return to chapter contents
Multi-site inference receipt
| Claim | Evidence | Boundary |
|---|---|---|
| Variables harmonised | Item and coding crosswalk | Construct equivalence |
| Designs comparable | Site diagnostics | Seed and fieldwork differences |
| Pooling justified | Estimand and weights | Named sites only |
| Transfer supported | Held-out site test | New population drift |
Hierarchical shrinkage is an assumption-bearing method, not automatic truth.
CHAPTER 9 OF 13 · Handle limits and disagreement
9. Handle cross-site mobility and duplicate eligibility
Core idea. Participants may belong to more than one site network, receive coupons across boundaries or enrol twice. Mobility can be substantive population structure, not merely data cleaning noise.
Method. Define residence and activity catchments, record referral origin, use privacy-preserving duplicate checks where authorised, model cross-site ties and publish how ambiguous cases affect estimands.
Worked example. A mobile participant works in City A and lives in City B. Excluding every cross-boundary referral would distort the real network; counting two enrolments would distort the sample.
Changed-context transfer. Changed-context transfer: apply the same reasoning structure to a school or public-service decision. Preserve the formal relationship, then replace the population, authority, privacy duties, evidence and remedy; never copy the domain conclusion.
Limits. Identity linkage can create privacy risk and may be legally restricted. Probabilistic matching produces its own uncertainty.
Evidence and review. For this chapter, keep a dated evidence record: the input, rule or assumption, result, accountable owner and a condition that would change the conclusion. Work an ordinary case and a boundary case, preserve rejected alternatives, and require an authorised independent reader to reproduce where judgement entered. Chapter 9 is complete only when the claimed result survives that review.
Misconception to avoid. Every cross-site record is fraud. Mobility and overlapping communities may be genuine features of the target population.
Practical evidence check
- Create duplicate, mobile and ambiguous cases and test the policy’s effect on each site denominator and chain.
- Return to chapter contents
CHAPTER 10 OF 13 · Handle limits and disagreement
10. Test transportability to a new site
Core idea. Transportability asks whether a relationship or prevalence model learned from named sites applies where population mix, network structure, recruitment implementation or calendar conditions differ.
Method. Describe effect modifiers available in both source and target, compare support, use leave-one-site-out validation, reweight only within overlap and declare an out-of-domain state.
Worked example. A model trained on three mature urban sites poorly predicts a smaller new site with sparse venues and high seed concentration; the system withholds a pooled prediction.
Changed-context transfer. Changed-context transfer: apply the same reasoning structure to a school or public-service decision. Preserve the formal relationship, then replace the population, authority, privacy duties, evidence and remedy; never copy the domain conclusion.
Limits. A handful of sites cannot reveal every modifier, and target data may be too limited to diagnose lack of overlap.
Evidence and review. For this chapter, keep a dated evidence record: the input, rule or assumption, result, accountable owner and a condition that would change the conclusion. Work an ordinary case and a boundary case, preserve rejected alternatives, and require an authorised independent reader to reproduce where judgement entered. Chapter 10 is complete only when the claimed result survives that review.
Misconception to avoid. If eligibility is identical, findings transport. Eligibility is only one part of population, network and implementation comparability.
Practical evidence check
- Hold out each site in turn and compare predicted with observed site estimates and diagnostic patterns.
- Return to chapter contents
CHAPTER 11 OF 13 · Handle limits and disagreement
11. Govern mid-course site adaptations
Core idea. Sites often need different seed additions, coupon rules, staffing or closure dates. Harmonisation should not prevent ethical adaptation, but changes must be versioned because they alter comparability and estimator validity.
Method. Use shared intervention categories, local justification, predeclared triggers, effective dates and analysis flags; rerun site and pooled sensitivity analyses before promoting post-change estimates.
Worked example. Site C adds venue-diverse seeds after a bottleneck while other sites do not. The pooled analysis retains pre/post phases and does not call the intervention ordinary recruitment.
Changed-context transfer. Changed-context transfer: apply the same reasoning structure to a school or public-service decision. Preserve the formal relationship, then replace the population, authority, privacy duties, evidence and remedy; never copy the domain conclusion.
Limits. Operational urgency can precede approval and small cells may prevent separate estimation. Governance must support timely, documented action.
Evidence and review. For this chapter, keep a dated evidence record: the input, rule or assumption, result, accountable owner and a condition that would change the conclusion. Work an ordinary case and a boundary case, preserve rejected alternatives, and require an authorised independent reader to reproduce where judgement entered. Chapter 11 is complete only when the claimed result survives that review.
Misconception to avoid. All sites must follow identical fieldwork to be comparable. Mechanically identical procedures can be inappropriate in different networks.
Practical evidence check
- Audit one adaptation from dashboard trigger through local decision, data flag, estimator check and public limitation.
- Return to chapter contents
CHAPTER 12 OF 13 · Transfer and govern
12. Connect multi-site estimates to responsible decisions
Core idea. Combined estimates may allocate services, compare communities or justify surveillance. Decision use must respect uncertainty, local autonomy, privacy and the possibility that pooling masks a harmed site.
Method. Publish local and combined views, use prediction intervals for transfer, conduct burden and benefit review, retain local interpretation and create correction and appeal routes.
Worked example. A regional programme uses pooled evidence for procurement but keeps site-specific outreach plans because recruitment and prevalence patterns differ materially.
Changed-context transfer. Changed-context transfer: apply the same reasoning structure to a school or public-service decision. Preserve the formal relationship, then replace the population, authority, privacy duties, evidence and remedy; never copy the domain conclusion.
Limits. Resource constraints can make bespoke decisions difficult, and public site comparisons can stigmatise communities.
Evidence and review. For this chapter, keep a dated evidence record: the input, rule or assumption, result, accountable owner and a condition that would change the conclusion. Work an ordinary case and a boundary case, preserve rejected alternatives, and require an authorised independent reader to reproduce where judgement entered. Chapter 12 is complete only when the claimed result survives that review.
Misconception to avoid. A single pooled number is fair because everyone is treated alike. Equal presentation can hide unequal evidence and consequences.
Practical evidence check
- Ask each site team to explain the combined claim, its local relevance and one reason to withhold it.
- Return to chapter contents
CHAPTER 13 OF 13 · Transfer and govern
13. Publish a multi-site RDS inference receipt
Core idea. A trustworthy receipt connects site populations, items, degree definitions, recruitment trees, diagnostics, local and cross-site weights, pooling model, heterogeneity, transportability, adaptations and correction history.
Method. Release de-identified synthetic trees, variable crosswalks, formulas, site panels, sensitivity code and decision logs; protect identities and vulnerable-community details through controlled review.
Worked example. An independent analyst reproduces one site estimate, three pooling schemes, one held-out-site prediction and the effect of excluding a seed-dominated chain.
Changed-context transfer. Changed-context transfer: apply the same reasoning structure to a school or public-service decision. Preserve the formal relationship, then replace the population, authority, privacy duties, evidence and remedy; never copy the domain conclusion.
Limits. No receipt turns nonprobability recruitment into a census or proves exchangeability of future sites. Claims remain scoped and revisable.
Evidence and review. For this chapter, keep a dated evidence record: the input, rule or assumption, result, accountable owner and a condition that would change the conclusion. Work an ordinary case and a boundary case, preserve rejected alternatives, and require an authorised independent reader to reproduce where judgement entered. Chapter 13 is complete only when the claimed result survives that review.
Misconception to avoid. A common protocol certifies comparability. Only accumulated design, implementation and outcome evidence can support the shared estimand.
Practical evidence check
- Rebuild the pooled headline from raw site receipts and deliberately change one degree definition to trigger its retirement.
- Return to chapter contents
Build the specialist decision record
For “The Gold Standard of Intelligence | RDS Multi-Site Harmonisation, Site-Specific Recruitment, Pooled Estimands, Transportability and Hierarchical Uncertainty”, begin with one bounded decision rather than an abstract claim. Name the affected people, data or institutions; the time window; the authority for acting; and the conclusion the method is allowed to support.
Preserve source material before interpretation. Record definitions, exclusions, transformations and the person responsible for each judgement. A reviewer should be able to separate an observed fact from an inferred result, a policy choice and a presentation shortcut.
Work one ordinary case in full. Show the input, every material rule or calculation, the output and the action that follows. Then write what the same evidence would mean under the simplest credible alternative method.
Keep an explicit unresolved state. Missing, protected, conflicting or unevaluated information must not become zero, false, consent or absence merely because a form expects one value.
Finish the record with a claim ledger. For every sentence that could drive action, identify its supporting source, inferential step, confidence or legal status, intended reader and expiry condition. This keeps descriptive evidence, formal consequence, professional judgement and institutional authority from collapsing into one impressive-looking conclusion.
Run the boundary-case audit
Select a case near the method’s domain edge: a missing input, a disputed definition, an endpoint, a sparse group, a resource limit or a legal deadline. Boundary cases reveal the contract that routine examples can conceal.
Change one assumption at a time and retain every output. Mark what stays stable, what reverses and what becomes undefined. If several assumptions move together, report that dependence instead of attributing the change to one convenient cause.
Search deliberately for a counterexample to the strongest claim. A method that works on the published illustration can still fail under a permitted input, an alternative grouping, a changed unit or an inaccessible evidence pathway.
Translate the counterexample into plain language: who would receive a different decision, what evidence would be hidden, and which safeguard or narrower claim repairs the problem.
Include an adversarial but plausible case. Let a missing group, ambiguous identifier, extreme weight, conflicting rule or urgent deadline interact with normal system incentives. The aim is not theatrical failure; it is to learn whether ordinary pressure makes the method conceal uncertainty, shift burden or reward a misleading shortcut.
Test independent reproduction
Give a reviewer the bounded inputs, public method contract and permitted access, but not the desired answer. Ask them to reproduce the result and identify every place where judgement, local knowledge or unavailable evidence was required.
Compare the reviewer’s intermediate states, not only the final number or label. Matching outputs can hide different denominators, inference paths, missing-data treatments or authorities that will diverge on the next case.
Where reproduction fails, classify the cause: ambiguous definition, undocumented transformation, incompatible implementation, inaccessible source, calculation error or legitimate alternative interpretation. Each class needs a different remedy.
Retain the disagreement and resolution. A corrected record should show what changed, why it changed and which prior outputs or decisions must be revisited.
Run the reproduction on a changed but structurally similar case after the first discrepancy is repaired. A method is not transferable merely because its authors can explain the original example. The new case tests whether the public contract carries the reasoning or whether hidden expertise is still doing essential work.
Design correction, challenge and remedy
Assign a visible route for correcting inputs, challenging assumptions and reviewing the application of the method. Technical accuracy without a usable remedy leaves the affected person carrying an error they cannot contest.
Separate urgent live-case repair from later system learning. A current decision may need pause, reassignment, fresh evidence or a human review before the organisation finishes its longer investigation.
Protect confidential and privileged material during review, but do not use confidentiality as a reason for unrecorded authority. Layer public explanation, protected oversight and individual reasons according to lawful access.
Verify that correction reaches cached results, derived records, dashboards and future decisions. Closing a complaint is not enough if the old conclusion continues to travel.
Measure remedy from the affected person’s position. A revised database row is useful, but it is incomplete when a hearing, service allocation, published claim or downstream model still relies on the old result. Record notification, restored opportunity, compensation or procedural repair where the governing framework provides it.
Set review and retirement triggers
Review after a change in definitions, population, law, source access, measurement scale, software, vocabulary or incentives. A stable formula can become invalid when the world or the meaning of its inputs changes.
Monitor leading signals as well as final outcomes: coverage gaps, timeouts, overrides, complaints, disagreement, missingness and boundary frequency. These often reveal failure before a headline metric moves.
Predefine thresholds for recalibration, suspension and retirement. Do not force a method to answer after its assumptions fail merely because replacing it is inconvenient.
At retirement, preserve the versioned contract, evidence, known limitations and migration decision. State which historic results remain interpretable and which require warning, restatement or withdrawal.
Schedule one prospective review before confidence fades or controversy forces it. Assign named owners for evidence, implementation and rights impact, and invite at least one reviewer outside the design team. The review should be able to narrow the claim, change the method or stop use without requiring proof of catastrophic failure.
Questions readers ask
Can multi-site RDS records simply be merged?
- No. The target population, local design adjustments and cross-site weights must be defined first.
Should every site use identical fieldwork?
- Core constructs should be harmonised, but ethical local adaptations may be necessary and must be versioned.
What does hierarchical pooling add?
- It can stabilise sparse site estimates under explicit exchangeability assumptions while retaining site effects.
Does a site indicator fix RDS bias?
- No. It cannot repair unknown inclusion processes, seed dependence or differential recruitment.
How is transportability tested?
- With shared effect modifiers, support checks and prospective or leave-one-site-out validation.
What should remain public?
- Site-specific evidence, pooled estimand, weights, heterogeneity, limits and correction history—without exposing participants.
Practise, verify and continue
Choose one consequential case. Reproduce the complete method, work one ordinary and one boundary case, test the strongest rival explanation, and state the evidence that would reverse or suspend the result. Build an evidence pack containing original inputs, definitions, calculations or rule steps, alternatives rejected, uncertainty, reviewer comments and the final action. Give the pack to a reader who did not design the method. Ask that reader to reproduce the result, identify one missing assumption and construct a case that would make the method return unresolved. Revise the public explanation only after the evidence pack and the operational process agree.
Move the method into an education or public-service setting. Keep the reasoning structure while replacing domain authority, population, privacy, evidence and correction routes. Start with a low-stakes historical example, then compare the specialist method with a transparent baseline. Check whether the same people, records and decisions remain inside scope; whether the scale or vocabulary still means the same thing; and whether a person affected by error can understand and challenge it. Pilot under human review, record reversals and burden, and stop the transfer if the formal resemblance hides a different ethical or legal problem.
Related eduKate reading
Authoritative sources and scope
CDC, Respondent-Driven Sampling Operations Manual
McKnight et al., Respondent-Driven Sampling in a Multi-Site Study
Gile and Handcock, Respondent-Driven Sampling: An Assessment of Current Methodology
Yauck et al., General regression methods for respondent-driven sampling data
WHO, Biobehavioural survey guidance for populations at risk
This article is sampling-method education, not a prevalence estimate or a claim that respondent-driven sampling identifies every target population. Examples and numbers are synthetic. Multi-site inference depends on eligibility, network boundaries, seeds, degree measurement, recruitment, timing, estimator assumptions, site weights, overlap and transportability. Protect participants and vulnerable communities, follow ethics and local governance, and use qualified survey-statistical review before consequential decisions.