From jom-skills
Runs and reports statistical analyses for JOM empirical manuscripts including measurement validity, endogeneity, and robustness checks for survey, archival, and experimental operations data.
How this skill is triggered — by the user, by Claude, or both
Slash command
/jom-skills:jom-data-analysisThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
- Operations data are collected and it is time to estimate and report
For survey-based OM constructs, defend the measurement model first:
| Operations data structure / claim | Estimator |
|---|---|
| Latent constructs, mediation, full survey model | SEM (covariance-based) or PLS-SEM where prediction/formative |
| Nested data (respondents in plants/firms) | Multilevel / HLM |
| Archival operations panel with unit heterogeneity | Fixed/random effects, high-dimensional FE; cluster-robust SE |
| Causal claim from secondary data | DiD/staggered DiD, IV/2SLS, matching, RD as the design fits |
| Count outcomes (recalls, defects, disruptions) | Poisson / negative binomial |
| Time-to-event (failure, project completion) | Cox / parametric survival |
| Manipulated operational decision | ANOVA/regression with manipulation & attention checks |
Cluster standard errors to the sampling/operational structure (plant, firm, supply tie).
Report the designed separations from jom-methods first (temporal/source/respondent separation), then statistical evidence: a Harman single-factor test is necessary but weak — prefer a marker variable, an unmeasured latent method factor, or showing interaction effects survive. Multi-respondent dyadic data is the strongest procedural remedy.
Recalls, supplier ties, lean adoption, and disruptions are rarely exogenous. State the threat (selection, reverse causality, omitted operational confounds), the identification strategy, and its assumptions. Report first-stage strength for IV and parallel-trends/anticipation checks for DiD.
The values below are widely cited conventions, not hard cutoffs; confirm field norms against current methods guidance.
| Diagnostic | Conventional landmark | Wanted alongside it |
|---|---|---|
| Alpha / composite reliability | typically ≥ .70 | source and prior validation of each scale |
| CFA fit (CFI/TLI) | typically ≥ .90–.95 | the model beating one-factor rivals |
| AVE (convergent) | commonly ≥ .50 | AVE > squared correlation, or HTMT |
| IV first-stage F | strong-instrument heuristics | why the instrument is excludable |
A study regresses plant defect rates on lean-adoption over a 9-year panel; adopters show 18% fewer defects (illustrative). A referee objects that plants adopting lean may already be better-managed, so selection contaminates the estimate. The JOM-grade response is an identification plan, not a footnote: exploit a staggered corporate mandate as quasi-exogenous timing, run a staggered DiD with plant and year fixed effects, cluster at the plant, and show pre-adoption parallel trends plus no anticipation. Report event-study coefficients so the dynamic effect is visible. If pre-trends are flat and the drop concentrates after the mandate, the inference is credible; if pre-trends slope, soften the claim to association.
【Measurement】alpha/CR, CFA fit, AVE/discriminant (survey) — pass/issues
【Estimator】SEM / HLM / panel-FE / DiD-IV / count / survival / experiment; SE clustering ...
【CMB / identification】designed separation + test; or endogeneity strategy + assumptions ...
【Mediation/Moderation】bootstrap CI / simple slopes reported? ...
【Robustness】...
【Next step】jom-contribution-framing
npx claudepluginhub brycewang-stanford/awesome-journal-skills --plugin jom-skillsRuns and reports statistical analysis for AMJ manuscripts: measurement validity (CFA, reliability), estimator selection (HLM, SEM, panel, experiments), common-method bias, endogeneity, and robustness checks.
Executes and stress-tests econometric, SEM/PLS, analytical-model, or ML analyses for JMIS manuscripts. Handles identification, endogeneity, construct validity, and robustness checks.
Runs and validates SEM/CFA, HLM/multilevel, regression, mediation/moderation, and meta-analytic estimation for JOM manuscripts. Use when estimation and results are the bottleneck.