| name | humrel-data-analysis |
| description | Use when executing and reporting the analysis for a Human Relations (HR) manuscript — qualitative coding and data-to-theory construction, critical interpretation, or quantitative estimation and robustness. Makes the evidence-to-theory link transparent; it does not design the study (see humrel-methods). |
Data Analysis & Evidence (humrel-data-analysis)
When to trigger
- You have data but the path from data to theory is opaque
- Qualitative: quotes are decorative, not evidentiary; coding is undocumented
- Critical: the interpretation reads as assertion rather than disciplined reading of the material
- Quantitative: main results exist but theorizing stops at the coefficient
- A reviewer asks "how did you get from your data to these constructs?"
The HR bar: make the inference auditable, then theorize beyond it
HR judges each tradition on its own terms, but every branch must satisfy the same demand: a reader should be able to see how the evidence became theory, and the analysis must yield the "unique and substantive theoretical contribution" the journal screens for. The relational, social nature of work should remain visible in the analysis — not abstracted away into variables or quotations stripped of context.
Branch A — Qualitative analysis (the data-to-theory ladder)
- Transparent coding. Document first-order codes (informant terms), second-order themes (your constructs), and aggregate dimensions — a Gioia-style data structure — or an equivalent (Eisenhardt cross-case, Langley process bracketing). State who coded, how disagreements were resolved, and how iteration proceeded.
- Data-to-theory table. Link representative raw evidence → codes → constructs so the inference is auditable (build with
humrel-tables-figures).
- Power quotes vs. proof quotes. A few vivid quotes in the body; corroborating quotes in tables/appendix. Quotes must carry the claim, not illustrate it after the fact.
- Patterned evidence + negative cases. Back each construct with evidence across informants; report disconfirming instances and how they refined the theory.
- Process display. For process theory, show the temporal/event structure (timeline, phase model, visual map).
Branch B — Critical analysis
- Make the interpretive procedure explicit (how texts/talk/practices were read; which discursive or material features mattered) so the reading is disciplined, not just asserted.
- Keep reflexivity active: how your standpoint shaped the interpretation.
- Tie the critique to a constructive theoretical claim — the analysis should leave readers with a new way to understand, not only a debunking.
Branch C — Quantitative analysis
- Main models match the design (multilevel/mixed, panel FE, SEM, event-history) with standard errors clustered at the right level.
- Construct validity in the analysis: report reliabilities, factor structure, and discriminant validity; address common-method concerns with design or statistical remedies where same-source.
- Robustness that targets the theory's threats: alternative measures, samples, specifications, and endogeneity checks — not a table farm.
- Interpret magnitudes in substantive, relational terms, not significance stars alone; HR house style for exhibits avoids decorating tables with asterisks as the "result."
- Probe the mechanism (mediation/moderation or supplementary tests), don't stop at X→Y.
Either branch — the "so what" of the evidence
- Tie every result back to the mechanism and the theoretical surprise.
- Distinguish what the data can and cannot establish — overclaiming is a fast route to rejection.
- Mind the data transparency matrix: if several papers draw on the same dataset, you must declare them and provide a matrix of which variables/quotations each uses; failure is grounds for rejection (检索于 2026-06；以官网为准).
Execution bridge (StatsPAI / Stata MCP)
Run the battery, don't just enumerate it. Full map:
execution-with-mcp. Human Relations blends critical/qualitative and quantitative work; apply the chain below to its survey / experimental quantitative papers.
- 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.
Checklist
Anti-patterns
- "Anecdotal" qualitative work: cherry-picked quotes, no coding transparency
- Quotes that illustrate a pre-set conclusion rather than supporting it
- Critical readings asserted with no statement of how the material was analyzed
- Robustness theater that never addresses the real threat
- Reporting significance with no substantive magnitude or relational meaning
- Concealing other papers built on the same dataset
Output format
【Journal】Human Relations
【Skill】humrel-data-analysis
【Branch】qualitative / critical / quantitative
【Data-to-theory link】data structure / interpretive procedure / mechanism tests
【Key evidence】power quotes or main estimates (with magnitude)
【Robustness/trustworthiness】checks done + gaps
【Transparency】same-dataset matrix needed? (yes/no/NA)
【Next skill】humrel-contribution-framing
