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
1---2name: jmr-data-analysis3description: Use when running and reporting the analysis for a Journal of Marketing Research (JMR) manuscript — selecting the estimator that matches the design, and meeting JMR's hard journal-level reporting mandate of exact p-values, standard errors, and effect sizes, plus replication-ready disclosure. Executes and reports; jmr-methods designs the study and jmr-contribution-framing states the payoff.4---56# Data Analysis & Reporting (jmr-data-analysis)78## When to trigger910- Data are collected (experimental or observational) and it is time to estimate and report11- You are unsure whether your estimator matches your design12- You must conform to JMR's exact-statistics reporting rules13- A reviewer says "the analysis does not support the inference" or "report effect sizes"1415## JMR's hard reporting mandate (journal-level)1617JMR enforces statistics reporting more explicitly than generic top journals. Empirical papers must report:1819- **Actual p-values to three digits** — *not* thresholds (no "p < .05"), *not* asterisks.20- **Standard errors** of parameter estimates in tables.21- **Effect sizes** — and a discussion of practical magnitude, not just significance.2223AMA 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.2425## Choose the estimator that matches the design2627| Design / claim | Estimator |28|-------------------------------------------------|-----------------------------------------------------------------|29| Experiment (factorial, between/within) | ANOVA / regression; estimated marginal means; planned contrasts |30| Behavioral mediation | Bootstrapped indirect effects (e.g., PROCESS), bias-corrected CIs |31| Moderation / moderated mediation | Interaction term + simple slopes; conditional indirect effects |32| Panel / observational causal | FE / DiD (modern staggered estimators); cluster-robust SE |33| Endogenous regressor | IV/2SLS, control function; report first stage and instrument tests |34| Discrete choice / demand | Logit/probit; random-coefficient (BLP-style) demand |35| Heterogeneity | Hierarchical Bayes / mixture models |36| Counts / limited DV | Poisson/NB, Tobit, as the outcome requires |3738Cluster standard errors to the sampling/assignment structure (e.g., by participant, store, or market).3940## Behavioral analysis specifics4142- Report manipulation- and attention-check results before the main effect.43- Mediation: bootstrap indirect effects with bias-corrected CIs (e.g., 5,000 resamples); for moderated mediation report the conditional indirect effect.44- Moderation: report the interaction coefficient, plot simple slopes, and give effect sizes per cell.4546## Modeling / econometric specifics4748- Report identification diagnostics (first-stage strength, parallel-trends/pre-trends, balance, overidentification) as relevant.49- Report structural parameter estimates with standard errors; show fit and counterfactuals where the contribution rests on them.5051## Result-to-claim ledger5253For each table or study, write one ledger row before drafting results:5455| Result | Claim it supports | Required statistic | Practical meaning |56| --- | --- | --- | --- |57| 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 |58| 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 |59| Robustness / alternative model | Which threat is reduced? | Same reporting discipline as main result | Whether conclusion changes in magnitude or direction |60| Counterfactual / simulation | What marketplace decision follows? | Parameter uncertainty and sensitivity | Managerial action implied by the estimate |6162If 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.6364## Replication & robustness (AMA transparency policy)6566- 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.67- Put robustness — alternative specifications, subsamples, alternative measures, additional studies — in the **'W'-prefixed Web Appendix**, keeping the print paper within 50 pages.6869## Execution bridge (StatsPAI / Stata MCP)7071Run the battery, don't just enumerate it. Full map:72[`execution-with-mcp`](../../../shared-resources/empirical-methods/execution-with-mcp.md). 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.7374- **Many outcomes / specifications:** `romano_wolf` (step-down FWER) or75 `benjamini_hochberg` — report the adjusted threshold.76- **OVB sensitivity:** `oster_delta` / `sensemakr`.77- **Inference:** `wild_cluster_bootstrap` (few clusters), `twoway_cluster` / `conley`;78 multilevel data → cluster at the right level.79- **Re-fit off one handle:** `audit_result(result_id)` lists the missing checks and the80 exact `suggest_function` for each.81- **Exhibits:** `etable` / `did_summary_to_latex` from the handle — no retyped numbers.8283Keep the decisive checks in the body and the exhaustive battery in the appendix. See the84executed chain in the [JF execution walkthrough](../../../Journal-of-Finance-Skills/resources/worked-examples/02-execution-walkthrough.md).85## Anti-patterns8687- Reporting "p < .05" or asterisks instead of exact three-digit p-values.88- Tables with no standard errors; significance without effect sizes.89- Causal-steps (Baron-Kenny) mediation instead of bootstrapped indirect effects.90- Ignoring clustering / non-independence; a weak or untested instrument.91- A leading zero before the decimal, or more than three decimal places.92- Results paragraphs that report significance but no practical magnitude or marketing interpretation.9394## Output format9596```text97[Target] JMR98[Genre] behavioral / modeling-econometric99[Estimator] matches design? SE clustering ...100[Exact stats] p three-digit / SEs / effect sizes: pass/fix101[AMA number style] no leading zero, <= 3 decimals: pass/fix102[Identification or process] diagnostics reported103[Result-to-claim ledger] claim + practical meaning complete104[Replication] Web Appendix + code/materials ready105[Next skill] jmr-contribution-framing106```107108## Resources109110- [`../../resources/official-source-map.md`](../../resources/official-source-map.md)111- [`../../resources/external_tools.md`](../../resources/external_tools.md)
brycewang-stanford