Use when running and reporting the statistical analysis for an Academy of Management Journal (AMJ) manuscript — measurement validity, common-method bias, the right estimator (HLM, SEM, panel, experiments), endogeneity, and robustness. Executes and reports the analysis; it does not design the study (amj-methods) or frame the contribution (amj-contribution-framing).
SkillScore100Maintained
Automatic
◆Data Analyst pack
Last verified 8/15/2026
What it changes
Passed 22 of 22 test cases with the skill, and 19 of 22 without it.
Without skill86%
With skill100%
Checks it passed that the model alone missed
The output uses the exact section header format 【Measurement】.
The output states amj-contribution-framing in the Next step section.
The output specifies amj-contribution-framing in the Next step section.
and 2 more in the full report
3 fixed19 passed either way0 still failing0 broke
Pass-rate lift
+14 pp
Tokens
+42%
Agent turns
0%
Response time
-15%
Model gemini-3.6-flashMeasured 8/15/2026v122 of 22 cases passed with skill
Tokens, turns and time describe resource use, not a dollar price.
See it in action
One of 3 test cases the skill fixed. Same model, same prompt, with and without the skill.
1 of 3
Prompt
I am preparing an empirical management manuscript for journal submission and need a thorough review of my data analysis and statistical reporting. Please evaluate my draft and format your output into sections with bracketed headers covering: Measurement (reliabilities, factor fit, construct validity), Estimator (model fit and standard error clustering), CMB evidence (procedural and statistical remedies), Mediation/Moderation (indirect effects and interaction testing), Endogeneity (identification strategy and assumptions), Robustness (alternative specifications), Open issues for reviewers, and Next step.
What was checked, without and with the skill
Checked
Without skill
With skill
The output uses the exact section header format 【Measurement】.
Missed
Passed
The output specifies amj-contribution-framing in the Next step section.
Missed
Passed
case-01 · gemini-3.6-flash · 8/15/2026 · answers are capped and redacted
What's inside
The instructions your model reads when the skill loads.
SKILL.md98 linessingle file
Data Analysis & Validity (amj-data-analysis)
When to trigger
Data are collected and it is time to estimate and report
You are unsure whether your estimator matches your design (nested data, latent constructs, panel)
Reviewers will probe measurement validity, common-method bias, or endogeneity
Interaction/mediation effects need correct testing and reporting
A reviewer says "the analysis does not support the inference" or "validity is not established"
Establish measurement before estimation
AMJ reviewers expect the measurement model to be defended first:
Reliability: Cronbach's alpha and/or composite reliability for each multi-item scale.
Confirmatory factor analysis (CFA): report fit (e.g., CFI, TLI, RMSEA, SRMR) and show the hypothesized factor structure fits better than plausible alternatives (one-factor, combined-factor models).
Convergent & discriminant validity: AVE per construct; AVE > inter-construct squared correlations (or HTMT). Report the correlation matrix with reliabilities on the diagonal.
Aggregation (multilevel): justify with ICC(1), ICC(2), and r_wg(j) before aggregating to a higher level.
Qualitative analysis: where the design is qualitative, "validity" becomes trustworthiness — present a Gioia-style data structure (first-order codes → second-order themes → aggregate dimensions), an audit trail, and representative quotations so the path from raw data to constructs is traceable.
Choose the estimator that matches the design
Data structure / claim
Estimator
Latent constructs, mediation, full model
Structural equation modeling (SEM)
Nested data (indiv. in teams/firms)
Multilevel / hierarchical linear modeling (HLM)
Panel with unit heterogeneity
Fixed/random effects; cluster-robust SE
Manipulated cause
ANOVA/regression with manipulation & attention checks
Poisson/negative binomial, logit/probit, Tobit as fits
Match clustering of standard errors to the sampling/nesting structure.
Common-method bias (CMB)
Report the designed separations from amj-methods first; then provide statistical evidence: a Harman single-factor test is necessary but weak — prefer a marker-variable approach, an unmeasured latent method factor (CFA), or showing interaction effects survive (interactions are hard to inflate by CMB). The Podsakoff et al. framework is the expected reference for both procedural and statistical remedies.
Reporting mediation, moderation, and effect sizes
Mediation: report indirect effects with bias-corrected bootstrap confidence intervals (e.g., 5,000 resamples); report the conditional indirect effect for moderated mediation.