技能 数据科学 系统综述方案与预注册

系统综述方案与预注册

v20260724
revedres-proposal-and-commissioning
本指南旨在帮助用户为教育研究的系统综述或荟萃分析制定严谨的、预先设定的研究协议(Protocol)。它指导使用者在实际搜索文献前,系统性地明确研究问题范围(PICO)、纳入排除标准、检索数据库和分析模型,从而确保研究流程的高度透明性和科学性。
获取技能
245 次下载
概览

Protocol & Preregistration (revedres-proposal-and-commissioning)

When to trigger

  • Topic selection cleared the four RER fit tests and it is time to lock the plan
  • You are about to search but have not written down inclusion/exclusion rules
  • You want a PROSPERO / OSF registration so the search is demonstrably a-priori
  • A reviewer is likely to ask "were these inclusion rules set before or after you saw the results?"

RER is SUBMITTED, not commissioned — the protocol replaces the commission

Annual Reviews invite their authors; RER reviews are submitted and peer-reviewed, so no editor pre-blesses your scope. The credibility that an invitation would have signaled must instead come from a transparent, a-priori protocol. For a systematic review or meta-analysis this is not optional polish — it is the difference between a synthesis and a fishing expedition. (RER occasionally publishes solicited analytic reviews of special topics, but the default and the bar is a strong unsolicited submission — 检索于 2026-06;以官网为准.)

The protocol: decisions to fix BEFORE you search

Write these down first; a reviewer who senses they were chosen after seeing results will distrust the whole review.

  1. Question in PICO/PICOS form. Population (learners/settings), Intervention/exposure or construct, Comparison, Outcomes, Study designs. This is the scope.
  2. Eligibility criteria. Inclusion and exclusion on population, constructs, designs, outcomes, language, publication type (peer-reviewed? grey literature? dissertations?), and date range — each with a reason.
  3. Information sources & search strategy. Databases (ERIC, PsycINFO, Education Source, Web of Science, Scopus, ProQuest Dissertations), grey-literature sources, and at least one full search string to be reported verbatim.
  4. Screening procedure. Two independent screeners at title/abstract and full-text; how disagreements are resolved; the reliability statistic you will report.
  5. Data extraction & coding scheme. What you will code from each study (design, sample, measures, effect sizes, moderators, risk-of-bias) and the codebook.
  6. Synthesis plan. Meta-analytic model (fixed/random/multilevel/robust variance), effect-size metric, planned moderator and sensitivity analyses, and publication-bias diagnostics — or the explicit qualitative/narrative synthesis method for a non-poolable literature.
  7. Risk-of-bias / quality appraisal tool chosen in advance (e.g. a design-appropriate instrument).

Preregistration: register the plan, then deviate transparently

  • Systematic reviews / meta-analyses: register the protocol on PROSPERO (health-adjacent) or OSF before screening; cite the registration in the manuscript. Use PRISMA-P as the protocol-reporting checklist.
  • A registration is a commitment, not a cage — you may change the plan, but you must flag and justify every deviation in the paper (planned vs. exploratory analyses kept distinct).
  • Purely conceptual/critical syntheses are not "registered" the same way, but they still need an explicit, written scope and search logic that the paper reports.

Why a-priori discipline pays off at RER specifically

A protocol is not bureaucratic overhead — it is what lets a submitted review be trusted the way an invited one is. Three concrete payoffs:

  • It pre-empts the "fishing expedition" objection. A reviewer who sees that inclusion rules, the model, and the moderators were fixed before screening cannot claim they were chosen to produce a clean result.
  • It makes the PRISMA flow honest. Counts flow from rules set in advance, so the diagram reports what the search found rather than what survived after-the-fact pruning.
  • It separates confirmatory from exploratory. Anything not in the protocol is labeled exploratory in the paper — which protects the headline result and still lets you report interesting surprises.

A common failure is treating the protocol as a draft of the paper. It is not prose; it is a set of decision rules a stranger could execute and reach your corpus. Write it as rules, register it, then let the search run against it.

Checklist

  • Question in PICO/PICOS; scope bounded the same way the search will be
  • Eligibility criteria written with a reason for each inclusion/exclusion rule
  • Databases + grey-literature sources listed; at least one full search string drafted
  • Dual independent screening defined; reliability statistic chosen
  • Coding scheme / codebook drafted before extraction
  • Synthesis plan fixed (meta-analytic model + moderators + bias diagnostics, or narrative method)
  • Risk-of-bias tool chosen a-priori
  • Registered on PROSPERO/OSF (or scope/search logic written for a conceptual synthesis)

Anti-patterns

  • Searching first, then writing inclusion rules to fit the studies you liked
  • "Random-effects, REML" chosen after seeing heterogeneity rather than in advance
  • No codebook — coders improvise, and reliability cannot be defended later
  • Treating registration as a formality, then silently changing the question or outcomes
  • Promising an exhaustive search but planning only one database

Output format

【Question (PICOS)】<population / intervention-construct / comparison / outcomes / designs>
【Eligibility】<key inclusion + exclusion rules, each with a reason>
【Sources + search】<databases + grey lit; one full search string drafted? Y/N>
【Screening】dual independent? Y/N; reliability stat = <kappa/agreement>
【Coding scheme】codebook drafted before extraction? Y/N
【Synthesis plan】<meta-analytic model + moderators + bias diagnostics | narrative method>
【Risk-of-bias tool】<instrument, chosen a-priori>
【Registration】PROSPERO/OSF id or "conceptual synthesis — scope written"
【Next step】→ revedres-literature-synthesis (run the search to a PRISMA flow)
信息
Category 数据科学
Name revedres-proposal-and-commissioning
版本 v20260724
大小 6.23KB
更新时间 2026-07-29
语言