From jmis-skills
Executes and stress-tests econometric, SEM/PLS, analytical-model, or ML analyses for JMIS manuscripts. Handles identification, endogeneity, construct validity, and robustness checks.
How this skill is triggered — by the user, by Claude, or both
Slash command
/jmis-skills:jmis-data-analysisThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
- A regression "works" but the identifying variation and threats are not pinned down
JMIS reviewers are method-literate across econometrics, psychometrics, analytical modeling, and data science. The bar is that the analysis credibly supports the verb in your claim.
JMIS values managerial relevance: report economic magnitude (elasticities, marginal effects, dollar value, lift), not only significance. Translate the headline coefficient into what it means for a firm, platform, or decision.
A referee writes "the result is not robust." A weak reply adds ten specifications and reports they are "all significant." A JMIS-grade reply names the threat each check defends against: a placebo on a period before the platform change rules out a secular trend; an alternative control group rules out a coincident shock; a sensitivity analysis (e.g., Oster-style bounds on selection) shows the estimate survives plausible unobserved confounding; an alternative measure of the construct rules out operationalization artifacts. Robustness is not a quantity of regressions; it is a mapping from each surviving threat to the check that kills it.
Suppose the main coefficient implies the ranking redesign cut marginal-seller retention by 6 percentage points. State it that way, then carry it to the platform decision: at the observed seller base that is roughly N exits per quarter and a Y% variety reduction (illustrative). A JMIS reader wants the economic magnitude and its managerial reading, not just p < 0.01.
On IT-value and platform papers, the most common first-round attack is "this could be reverse causality or selection." Do not wait for it — pre-empt it in the analysis. Show the timing (the cause precedes the effect), use within-unit variation that differences out fixed selection, lean on a quasi-experimental shock where you have one, and where you must use an instrument, defend the exclusion restriction on institutional grounds and report weak-IV-robust inference if the first stage is not strong. Then bound what remains: a selection-sensitivity analysis (e.g., Oster-style δ/bounds) tells a referee how much unobserved confounding it would take to overturn the result. Anticipating the endogeneity objection inside the paper is worth more than answering it in a rebuttal.
JMIS submissions are double-anonymized and capped at 50 pages, which shapes how you present analysis. Online appendixes are permitted, but the body must stand on its own: a reviewer should be able to follow the identification and the headline result without the appendix, and should never find that a load-bearing robustness check exists only there. Document the sample construction, the estimator, the software/version, and the inference choices clearly enough that the result is traceable; if you reference your own prior code or data, phrase it so it does not de-anonymize you. Tighten the analysis narrative — full diagnostic batteries and secondary specifications belong in the appendix, while the body carries the chain that establishes the contribution.
Do not move to jmis-contribution-framing while the headline number is still drifting across specifications. The contribution sentence and the exhibits both depend on a settled effect size and a settled inference, so lock the preferred specification, confirm it survives the robustness battery, and fix the economic-magnitude interpretation before framing the claim. A contribution built on a coefficient that later moves forces a rewrite of the intro, the discussion, and the abstract.
【Evidence type】firm/platform econometrics / survey-SEM / experiment / analytical / ML
【Identification or validity】variation source + diagnostics / measurement model + CMB test / proof + robustness
【Inference】clustering / weak-IV-robust / bootstrap as applicable
【Magnitude】economic/managerial interpretation of the headline effect
【Robustness done】spec / sample / functional-form / out-of-sample
【Next step】jmis-contribution-framing
npx claudepluginhub brycewang-stanford/awesome-journal-skills --plugin jmis-skillsRuns and validates SEM/CFA, HLM/multilevel, regression, mediation/moderation, and meta-analytic estimation for JOM manuscripts. Use when estimation and results are the bottleneck.
Selects and defends JMIS-compatible research designs—IT-value econometrics, behavioral experiments, analytical models, or design-science artifacts—matching method to causal claim and page budget.
Runs and reports empirical analysis for JAIS manuscripts: SEM measurement/structural models, causal identification, artifact evaluation, or qualitative data structure.