jcr-data-analysis
GitHub用于JCR稿件的证据分析与报告，涵盖行为实验的中介/调节效应及CCT工作的可信度解释。执行引导分析并生成符合JCR透明度要求的报告，但不涉及研究设计。
触发场景
安装
npx skills add brycewang-stanford/Awesome-Journal-Skills --skill jcr-data-analysis -g -y
SKILL.md
Frontmatter
{
"name": "jcr-data-analysis",
"description": "Use when running and reporting the evidence for a Journal of Consumer Research (JCR) manuscript — process evidence (mediation\/moderation) for behavioral experiments, or trustworthy interpretation for Consumer Culture Theory (CCT) work — plus robustness and the transparency reporting JCR requires. Executes and reports; it does not design the studies (jcr-methods)."
}
Data Analysis & Evidence (jcr-data-analysis)
When to trigger
- Studies are run and it is time to analyze and report
- Reviewers will probe whether the data actually support the process claim
- You need bootstrapped indirect effects, simple-slopes, or spotlight/floodlight analyses
- You are reporting interpretive (CCT) evidence and need to defend trustworthiness
Process evidence is the JCR currency (experiments)
For the dominant experimental tradition, JCR reviewers ask whether the data support the psychological process, not just the effect:
- Mediation: report indirect effects with bias-corrected bootstrap confidence intervals (e.g., 5,000 resamples). Treat measured-mediator mediation as suggestive; moderation-of-process and manipulated-mediator designs are stronger and expected for a clean process claim.
- Moderation: report the interaction term, then plot simple slopes with regions of significance; for continuous moderators use spotlight/floodlight analysis rather than median splits.
- Effect sizes: report standardized effects (Cohen's d, eta-squared, or equivalents) and discuss practical magnitude, not just p-values.
- Across studies: show the effect converges across populations, stimuli, and operationalizations; a small internal meta-analysis of the effect can strengthen the package.
Reporting standards and clean inference
- Pre-specify and report exclusion rules, manipulation/attention-check pass rates, and final cell sizes; report all measured conditions and dependent variables (no selective reporting).
- Confirm random assignment worked (baseline balance) and that the manipulation moved the proposed mediator.
- Rule out the rival accounts named in
jcr-theory-developmentwith direct tests, not hand-waving.
Interpretive (CCT) evidence: trustworthiness, not p-values
Where the design is interpretive, "analysis" means a defensible path from raw data to conceptual claims:
- Present a traceable interpretation: representative quotations/observations linked to second-order constructs and an audit trail.
- Demonstrate triangulation, prolonged engagement, and member checks; show the conceptual framework is grounded in the corpus.
- Make the interpretive logic transparent enough that a reader could follow how the data yielded the theory.
Transparency reporting (JCR-specific)
JCR's research-transparency regime governs reporting: prepare the Data Collection Statement (where/when/who collected and analyzed the data, where stored — hidden from reviewers, published if accepted), post data and materials to an approved repository (OSF, Harvard Dataverse, Qualitative Data Repository, ResearchBox) as required at invited revision, supply statistical/programming replication code (or a written description for proprietary code), and plan to retain data ≥ 7 years. Consult JCR's "Research Method Transparency Guidelines and Reporting Requirements."
Execution bridge (StatsPAI / Stata MCP)
Run the battery, don't just enumerate it. Full map:
execution-with-mcp. JCR is predominantly lab experiments; randomization-based inference and the many-outcome family-wise correction (romano_wolf) are the decisive tools.
- Many outcomes / specifications:
romano_wolf(step-down FWER) orbenjamini_hochberg— report the adjusted threshold. - OVB sensitivity:
oster_delta/sensemakr. - Inference:
wild_cluster_bootstrap(few clusters),twoway_cluster/conley; multilevel data → cluster at the right level. - Re-fit off one handle:
audit_result(result_id)lists the missing checks and the exactsuggest_functionfor each. - Exhibits:
etable/did_summary_to_latexfrom the handle — no retyped numbers.
Keep the decisive checks in the body and the exhaustive battery in the appendix. See the executed chain in the JF execution walkthrough.
Checklist
- Mediation via bootstrapped indirect effects; process manipulated where possible
- Moderation via simple slopes / spotlight-floodlight, not median splits
- Effect sizes reported with practical interpretation
- Exclusions, check pass rates, and all conditions/DVs reported
- Rival accounts tested directly
- CCT: triangulation, member checks, audit trail, grounded framework
- Data Collection Statement, repository deposit, and replication code prepared
Anti-patterns
- Causal-steps "mediation" with no bootstrapped indirect effect.
- Median-splitting a continuous moderator instead of spotlight analysis.
- p-values with no effect sizes or practical meaning.
- Selective reporting of conditions, measures, or studies.
- CCT claims with no audit trail linking data to constructs.
Output format
【Genre】experiments / CCT
【Effect】direction, size, convergence across studies
【Process evidence】mediation (bootstrap CI) / manipulated mediator / moderation-of-process
【Moderation】simple slopes / spotlight-floodlight
【Rivals ruled out】[...]
【CCT trustworthiness】triangulation / member checks / audit trail
【Transparency】Data Collection Statement + repository + code: ready?
【Next step】jcr-contribution-framing
版本历史
- 1839142 当前 2026-07-05 13:30
