Jmr data analysis

Data Analysis & Reporting (jmr-data-analysis)

When to trigger

• Data are collected (experimental or observational) and it is time to estimate and report
• You are unsure whether your estimator matches your design
• You must conform to JMR's exact-statistics reporting rules
• A reviewer says "the analysis does not support the inference" or "report effect sizes"

JMR's hard reporting mandate (journal-level)

JMR enforces statistics reporting more explicitly than generic top journals. Empirical papers must report:

• Actual p-values to three digits — not thresholds (no "p < .05"), not asterisks.
• Standard errors of parameter estimates in tables.
• Effect sizes — and a discussion of practical magnitude, not just significance.

AMA results-reporting style: no leading zero before the decimal (write .97, p = .032), and no more than three decimal places. Apply this to every table and in-text statistic.

Choose the estimator that matches the design

Design / claim	Estimator
Experiment (factorial, between/within)	ANOVA / regression; estimated marginal means; planned contrasts
Behavioral mediation	Bootstrapped indirect effects (e.g., PROCESS), bias-corrected CIs
Moderation / moderated mediation	Interaction term + simple slopes; conditional indirect effects
Panel / observational causal	FE / DiD (modern staggered estimators); cluster-robust SE
Endogenous regressor	IV/2SLS, control function; report first stage and instrument tests
Discrete choice / demand	Logit/probit; random-coefficient (BLP-style) demand
Heterogeneity	Hierarchical Bayes / mixture models
Counts / limited DV	Poisson/NB, Tobit, as the outcome requires

Cluster standard errors to the sampling/assignment structure (e.g., by participant, store, or market).

Behavioral analysis specifics

• Report manipulation- and attention-check results before the main effect.
• Mediation: bootstrap indirect effects with bias-corrected CIs (e.g., 5,000 resamples); for moderated mediation report the conditional indirect effect.
• Moderation: report the interaction coefficient, plot simple slopes, and give effect sizes per cell.

Modeling / econometric specifics

• Report identification diagnostics (first-stage strength, parallel-trends/pre-trends, balance, overidentification) as relevant.
• Report structural parameter estimates with standard errors; show fit and counterfactuals where the contribution rests on them.

Result-to-claim ledger

For each table or study, write one ledger row before drafting results:

Result	Claim it supports	Required statistic	Practical meaning
Main treatment or model estimate	What marketing decision, mechanism, or theory point changes?	Exact p-value, standard error, CI/effect size	Unit change, percentage lift, WTP/profit/customer impact
Mediation/process result	Which mechanism is supported and which rival is weaker?	Indirect effect with CI; moderation where relevant	Why the process matters for managers or theory
Robustness / alternative model	Which threat is reduced?	Same reporting discipline as main result	Whether conclusion changes in magnitude or direction
Counterfactual / simulation	What marketplace decision follows?	Parameter uncertainty and sensitivity	Managerial action implied by the estimate

If the practical-meaning column is empty, the result is not ready for a JMR results paragraph. JMR reviewers expect precision, but they also expect a marketing payoff.

Replication & robustness (AMA transparency policy)

• Provide enough detail (in-text, Web Appendix, or online supplements) for a reasonably trained researcher to replicate; be ready to share code, instruments/stimuli, and materials on request, and to provide data/materials before final acceptance.
• Put robustness — alternative specifications, subsamples, alternative measures, additional studies — in the 'W'-prefixed Web Appendix, keeping the print paper within 50 pages.

Execution bridge (StatsPAI / Stata MCP)

Run the battery, don't just enumerate it. Full map: execution-with-mcp. JMR mixes experiments, structural models, and quasi-experiments; the chain below serves the experimental and reduced-form lanes, while structural demand estimation uses its own toolkit.

• Many outcomes / specifications: romano_wolf (step-down FWER) or benjamini_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 exact suggest_function for each.
• Exhibits: etable / did_summary_to_latex from 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.

Anti-patterns

• Reporting "p < .05" or asterisks instead of exact three-digit p-values.
• Tables with no standard errors; significance without effect sizes.
• Causal-steps (Baron-Kenny) mediation instead of bootstrapped indirect effects.
• Ignoring clustering / non-independence; a weak or untested instrument.
• A leading zero before the decimal, or more than three decimal places.
• Results paragraphs that report significance but no practical magnitude or marketing interpretation.

Output format

[Target] JMR
[Genre] behavioral / modeling-econometric
[Estimator] matches design? SE clustering ...
[Exact stats] p three-digit / SEs / effect sizes: pass/fix
[AMA number style] no leading zero, <= 3 decimals: pass/fix
[Identification or process] diagnostics reported
[Result-to-claim ledger] claim + practical meaning complete
[Replication] Web Appendix + code/materials ready
[Next skill] jmr-contribution-framing

Resources

• ../../resources/official-source-map.md
• ../../resources/external_tools.md