RMP.7 - Synthesize a Research Evidence Corpus

Type: Research Method Practice practitioner method pattern

RMP.7:1 - Problem Frame

Use this pattern when several research reports, source interpretations or an existing synthesis bear on a question, and someone needs to determine what their combined contribution warrants. Start by stating the question the synthesis must answer and checking what the supplied corpus actually contains. Reuse an adequate existing synthesis.

Four reports can represent only two independent tests plus a different kind of material. Several historical accounts can repeat one memoir. Studies can disagree because their questions or settings differ. The useful first result separates these relations before treating apparent agreement as stronger support.

This Method produces an EvidenceSynthesis: the warranted corpus answer, how the contributions were related, and the limits needed for criticism and reuse. A short answer linked to an existing review may suffice. A new analysis may need a study map, comparison or computation; the result name requires no universal database or dossier.

Enter directly with a qualified question, applicable synthesis design, operational choices and trace adequate for this use, whether supplied or developed through RMP.1–RMP.4. A design may permit reasoned refinement as historical or qualitative material becomes understood. Preserve the resulting question and criticism, or explicitly return the changed premise.

Use the relevant source Method when finding or comparing sources already closes the question. Use RMP.5 for a particular material-to-claim inference. RMP.7 adds the research-specific difficulty of appraisal and integration across a corpus. A missing disciplinary synthesis Method, inaccessible decisive material or unsettled question returns to its supplier with the affected conclusion identified.

RMP.7:2 - Problem

A collection can look like a body of corroborating evidence while sharing one underlying study, selecting favorable results or combining incompatible claims. A pooled number can conceal that its inputs measure different things. A common theme can conceal a consequential difference in setting or meaning.

The opposite response is to abandon every synthesis that contains disagreement or imperfect sources. That discards compatible quantitative combination, a useful interpretive account or a precise explanation of what the corpus cannot establish. Another unnecessary cost is recreating an adequate review merely to produce a new dated record.

The practitioner needs a corpus answer whose form follows the research question, the source relations and the criticism that matters.

RMP.7:3 - Forces

RMP.7:4 - Solution

Recover the question and actual corpus, distinguish the contributions and their dependence, appraise the inference each can support, and perform the justified synthesis. Return the answer with its consequential transformations, uncertainty and scope.

RMP.7:4.1 - State the question and test the supplied answer

Name the subject, comparison or interpretive concern, relevant context and period, and the kind of conclusion wanted. A mean difference, explanation, account of experiences and claim about a historical sequence ask different questions even when they concern the same topic.

Inspect any supplied synthesis for the same question, relevant source coverage, Method, actual findings, limitations and current receiving use. If it already supplies an adequate answer, return that answer with the source and qualifications that change reliance. Reopen only a changed premise or uncovered question.

Otherwise establish the actual source set against the selected design: what was eligible, sought, obtained, excluded or unavailable, and why those choices bear on this answer. Systematic-coverage claims need the corresponding search and selection work. A deliberately bounded corpus can answer its bounded question without claiming to represent every study.

Return a changed question to RMP.1, an unsettled synthesis design to RMP.2, an operational problem to RMP.3 or an actual-material problem to RMP.4. An adequate supplied input needs no recreation of those earlier activities.

RMP.7:4.2 - Recover distinct contributions before counting support

Distinguish a report from its underlying studies and material, a source author’s interpretation from the material it interprets, and either from the proposed synthesis finding.

Keep contrary findings and disagreements between reports visible. If dependence cannot be recovered well enough for the intended combination, qualify or withhold that combination and state what comparison remains possible.

RMP.7:4.3 - Appraise support for the proposed findings

Use the criteria appropriate to the research tradition and each finding. Distinguish limitations within a study, dependence among contributions, selective availability across the corpus and applicability to the synthesis question. Their effects can differ across outcomes and interpretations.

Examine consequential exclusions and missing material. Distinguish evidence that was never sought, an eligible result not reported, a lost record and a source that cannot be accessed. Identify which conclusion might change; absence from the available corpus alone does not establish an absent effect, event or experience.

A large number of publications cannot repair common dependence. A global rating for each source cannot substitute for the relation that warrants an inference. Some weaker material may still provide context or challenge an interpretation; keep its supported role and limits explicit.

An individual claim defeated by its material or Method returns to RMP.5. Use RMP.6 when a remaining credibility choice can change the present synthesis or a worthwhile next move. Selecting an examination supplies no result from an examination not yet performed.

RMP.7:4.4 - Perform the synthesis the question permits

Apply a qualified disciplinary Method. Explain consequential grouping, transformation, weighting or translation so that another reader can recover how source contributions became the finding.

Variation does not automatically prohibit pooling; the question and model determine what variation means. Conversely, choosing a random-effects label does not make different targets compatible. A direction-only synthesis cannot estimate an effect’s magnitude.

For qualitative synthesis, a participant’s account, the source author’s interpretation and a new synthesis construct retain their distinct contributions. When GRADE-CERQual fits the selected use, judge each qualitative finding through methodological limitations, coherence, adequacy and relevance. Its confidence terms are not numerical truth probabilities or a universal historical scale.

If the needed Method or a decisive premise is unavailable, return that specific gap and the bounded result still warranted. A polished common format cannot supply the missing inference.

RMP.7:4.5 - Return the useful corpus answer

State what the corpus establishes first. Include the question, material and Method actually used, consequential dependence and transformations, uncertainty, missing contributions and transfer limits that change interpretation.

Use existing source notes, an analysis output or a concise linked explanation when adequate. Preserve access conditions and enough connection to source material for the named criticism. A public synthesis and a critic’s permitted access may support different examinations.

Finish with the adequate existing synthesis, the produced finding, a qualified corpus answer, an informative refusal, or a precise input return. An unanswered broader question can remain visible. Further inquiry earns selection through its attainable contribution and whole burden; insufficient evidence remains insufficient when no additional work is selected.

Use RMP.8 when the answer requires reconciliation with a wider claim or model, or a return to an actual receiving use with consequential reliance conditions. An already sufficient answer can be returned directly.

RMP.7:5 - Archetypal Grounding

These constructed cases illustrate the Method; they report no performed study.

RMP.7:5.1 - Two compatible tests among four reports

A seal engineer receives reports A, A2, B and C. A/A2 describe the same randomized accelerated test; A2 supplies its fuller outcome account. B describes an independent test of the same material contrast under compatible stress and measurement conditions. C describes in-service replacement choices under different conditions.

The synthesis question is the difference in mean time to the defined laboratory failure between the new and comparison seal under the specified accelerated conditions. Assume complete observed lifetimes, suitable approximately normal estimates, independent study estimates and a domain-justified common effect. The four-report map retains A2’s useful information without adding another independent estimate; C cannot estimate this laboratory contrast.

The recovered study estimates are 14 hours with standard error 4 for A/A2, and 8 hours with standard error 3 for B. Inverse-variance weights are 1/16 and 1/9. Their weighted estimate is 10.16 hours; its standard error is 2.4 hours, giving an approximate normal 95% interval of 5.46 to 14.86 hours.

The returned answer is a positive mean difference under the named laboratory conditions and assumptions. The study-specific estimates remain visible. The synthesis does not establish that the acceleration relation transfers to service or that a maintenance interval should change. C remains useful material for a different service-use question. An inadequate common-effect premise would reopen this calculation’s interpretation, even if the arithmetic were correct.

RMP.7:5.2 - A refusal with useful content

Another supplied corpus reports leakage after one hour, failure time under another stress and technicians’ replacement decisions. The requested answer is a common percentage reduction in 500-hour failure risk.

RMP.7 shows each result with its actual outcome, contrast, conditions and uncertainty. No supplied relation converts them to the requested risk. The result therefore says that this corpus does not establish that percentage and preserves the narrower comparisons it does support. It supplies neither a pooled number nor a conclusion of no effect. If the receiver asked only whether this corpus establishes the percentage, that question is complete.

RMP.7:5.3 - Situated participation and a mixed question

A resident group asks what impedes participation in water monitoring. One interview study links distrust to withheld feedback. Another interpretive study relates continued participation to community authority over the question. A survey reports stated reasons within its own sampled group.

Under a justified same-question mixed design, the synthesist retains the survey’s counts and denominator, the qualitative contexts and the source authors’ interpretations. A comparison suggests that control over both the question and returned information may matter. That explanatory connection is the synthesist’s contribution; it is not a measured population prevalence. A contrary case with continued participation despite withheld feedback qualifies the explanation.

For a separate question about the monitoring intervention’s effect on water quality, these accounts and any effect studies need their own syntheses before their relationship is examined. Agreement about participation is not an estimated effect on water quality.

RMP.7:5.4 - Shared memoir and contrary chronology

Several later histories attribute a factory’s decision to close to a public protest. Their references lead to one retrospective memoir. An authenticated record dates that decision before the protest.

The historian separates the shared evidential source from the later interpretations and examines what the decision record establishes. The protest cannot have caused that earlier decision. The decision’s cause and any effect on later implementation remain unresolved. The chronology is a useful completed corpus answer; counting the later histories would have concealed its decisive contradiction.

RMP.7:6 - Bias-Annotation

The numerical example makes statistical combination unusually easy: compatible targets, recovered dependence and justified assumptions are supplied. Real corpora may require domain expertise, additional source recovery or a narrower answer. The historical and participation cases expose different criteria; neither makes this pattern an exhaustive handbook for those traditions.

Publication and access practices can privilege certain institutions, languages or surviving sources. Interpret silence and corpus coverage in that setting. Reusing a convenient synthesis remains appropriate only for the question and qualifications it actually covers.

RMP.7:7 - Conformance Checklist

Use these questions to examine a proposed synthesis after recognizing the corpus difficulty:

• Can the receiver recover the exact question and why an existing answer does or does not suffice?
• Are reports, underlying studies or material, source interpretations and synthesis findings distinguishable, including consequential dependence?
• Do eligibility, exclusions, missing material and access limits support the stated corpus scope?
• Does each consequential transformation have a Method and an interpretation that fit its question?
• Are contrary findings, uncertainty and transfer limits retained at the findings they affect?
• Does the answer finish the bounded use or return the specific missing result without disguising an unperformed inquiry as evidence?

Recognition selects this Method. These questions examine the resulting synthesis; recognition alone does not establish its adequacy.

RMP.7:8 - Common Anti-Patterns and How to Avoid Them

RMP.7:9 - Consequences

The practitioner can return a defensible corpus answer while preserving useful differences between sources. Compatible contributions can strengthen an estimate; unlike or dependent contributions can instead change its scope, explanation or limits.

The cost is question-specific appraisal and explicit consequential transformations. Adequate-synthesis reuse and proportionate records reduce work without turning incomplete evidence into a stronger claim.

RMP.7:10 - Rationale

A source collection supplies material; the research synthesis establishes a further relation among its contributions. That relation needs its own design and criticism. Keeping it distinct from RMP.5’s individual inference makes dependence, coverage and integration visible without making synthesis compulsory after every analysis.

A useful refusal is informative when it identifies the missing relation and preserves the supported answer. It protects both legitimate combination and honest completion.

RMP.7:11 - SoTA-Echoing

What counts as independent support? The selected line recovers the underlying studies and source relationships while retaining additional information from their reports. The cheap default is to count publications or discard all apparent duplicates. The former inflates support; the latter can lose follow-up detail. Adapt: 4.2 combines information at the appropriate contribution and preserves consequential overlap, as illustrated by A/A2 in 5.1. Cochrane chapter 5 supplies the best-known-line candidate for these report/study distinctions because it resolves both failures, not because it is an institutional handbook. This recovery adds work when relationships are unclear; it is needed only where they can change the finding. New overlap, provenance or supplementary results reopen the affected contribution.

Which numerical answer can this corpus support? The selected line chooses a synthesis for a stated quantity and dependence model, with explicit alternatives when that quantity is unsupported. The serious alternatives are pooling every numerical result or refusing all heterogeneous material. The first invents comparability; the second can discard a warranted combined estimate. Adapt: 4.4 and 5.1–5.2 preserve justified combination and an informative narrower answer. Cochrane chapter 10 supplies the quantitative best-known-line candidate; chapter 12 supplies the competing non-meta-analysis route when its conditions fit. Their substantive question and interpretation rules justify the choice; intervention-review conventions do not govern every research family. The trade-off is model and grouping judgment in exchange for an interpretable answer. A changed target, dependence relation or defensible alternative model reopens that choice.

How should missing contributions affect a finding? The selected line locates what is missing and the particular inference it can change. Treating an available corpus as complete is cheaper but can hide selective results; rejecting every incomplete corpus would also erase bounded useful answers. Adapt: 4.1 and 4.3 retain question-relative coverage and distinguish unavailable, unreported, unsought and lost material. Cochrane chapter 13 is a best-known-line candidate for these missing-evidence questions; its meta-analysis tools retain that scope. New eligible material, an identified selective omission or a stronger model of its effect reopens the affected finding.

How can situated and mixed findings be related? The selected line preserves context and finding-specific confidence, integrating transformed contributions only when they answer a shared question. The serious rival is separate synthesis of all quantitative and qualitative contributions; a different default would merge everything into one theme. Cochrane chapter 21 supplies a best-known qualitative line and the separate-contribution choice; JBI §8.3 supplies the serious question-relative integration alternative. Adapt: 4.4 and 5.3 select between them by the question. This permits a useful shared answer while keeping different aspects separate and transformation losses recoverable. GRADE-CERQual supplies an applicable finding-level confidence candidate, not a numerical truth scale. The accepted cost is explicit interpretation and method choice. A changed tradition, question or stronger account of what transformation loses reopens the comparison.

What can historical agreement establish? The selected line compares source dependence, dating, purpose, contrary material and rival explanations. Narrative count is an attractive cheap default, but copied accounts cannot supply independent causal support. Adapt: 4.2–4.4 and 5.4 recover the shared memoir and the different contribution of the decision record. AHA professional standards supplies the best-known-line candidate for source integrity and evidence/interpretation discipline here; it supplies no verdict about the particular factory. The extra source criticism preserves the useful chronology while leaving the unsupported cause open. New provenance, dating or a substantive rival explanation reopens that inference.

These comparisons support different research traditions. Their combination here is question-relative guidance, not evidence that one synthesis procedure works universally.

RMP.7:12 - Relations

• RMP.1–RMP.4 can supply the qualified question, design, operations and actual trace. Compatible direct inputs are equally usable; return only a missing or changed premise.
• RMP.5 supplies examined source claims and receives a defeated individual inference. RMP.6 supplies an actual credibility result when available or decides a useful further examination.
• RMP.8 receives the synthesis when a wider claim or model must be revised or evidence returned to an actual receiving use. A sufficient bounded corpus answer can finish directly.
• FPF F.1, F.0.1 and A.10, with a qualified F.0.2 contribution when applicable, supply source selection, comparison and evidence-use relations. Their source work does not by itself establish this research synthesis.
• Domain, access and receiving-practice Methods supply their criteria, permissions and decisions. Research-Method improvement remains a distinct RMP.9/Method Engineering question.

RMP.7:End