Research Design & Methods (hrm-methods)

When to trigger

• Predictor and outcome are single-source, single-wave, self-reported (common-method-bias risk)
• The theory is cross-level (unit HR system → individual outcome) but data are one level
• HR-system or practice constructs lack validated measures or a defended aggregation
• The causal claim ("HPWS raises performance") rests on cross-sectional correlation
• A reviewer says "the design cannot test this hypothesis," "CMB," or "HPWS adoption is endogenous"

HRM accepts any rigorous design — the bar is fit, not method

HRM welcomes qualitative, quantitative, meta-analytic, and critical-review work, exploratory or confirmatory, inductive/deductive/abductive. The judgment is fit and rigor, plus the journal's demand that the design support a practice-relevant conclusion. Match the design to the claim:

A two-study design (e.g., a field study for generalizability plus an experiment for the mechanism) is a recognized HRM strength.

Designing against the threats HRM referees punish

• Common-method bias (CMB). Procedural remedies beat statistical fixes: separate the source (self-report predictor, supervisor/objective outcome) and the time (multi-wave lag) of measurement, and plan this before data collection. A Harman single-factor test or a single unmeasured-latent-method-factor model is a supplement, not a defense.
• Endogeneity of HR adoption (archival/strategic HRM). Firms choose HPWS for reasons correlated with performance. Anticipate omitted variables, reverse causality, and selection; specify an identification strategy (panel fixed effects, DiD, instrument, or natural experiment) and state the assumption each requires.
• Construct validity. Use validated multi-item scales; pilot new HR-practice measures; plan a CFA establishing convergent/discriminant validity. Be explicit whether you measure the intended, implemented, or perceived HR system — they are different constructs and require different respondents.
• Aggregation. When measuring a unit-level HR system from individual reports, justify aggregation with r_wg, ICC(1) / ICC(2), and a substantive composition model (referent-shift vs. direct consensus). Do not aggregate without a theory of why the construct is shared.
• Sampling and power. Justify the sampling frame and response rate; power the interaction/cross-level effects, which need more power than main effects.

Level-of-analysis discipline

State the level for theory, measurement, and analysis, and keep them aligned. If theory is unit-level but data are individual, justify aggregation; if effects are cross-level, the analysis must model the nesting — do not run OLS on nested data.

Execution bridge (StatsPAI / Stata MCP)

For the empirical / causal lane, estimate and audit rather than only specify. Full map: execution-with-mcp. HRM is empirical HR — multilevel survey data, field experiments, and panels; multilevel inference and many-outcome corrections matter most.

• detect_design → recommend → fit with as_handle=true → audit_result to enumerate the checks the design owes.
• Panel / staggered DiD: callaway_santanna / sun_abraham + bacon_decomposition
  • honest_did_from_result. IV: effective_f_test + anderson_rubin_ci. RDD: rdrobust + mccrary_test.
• Experiments: randomization-based inference and romano_wolf for the many-outcome family-wise correction reviewers expect.

Match the toolchain to the reviewer pool, and report the effect size the venue wants. A run end-to-end (synthetic data, real returns) is in the JF execution walkthrough.

Checklist

• Design can actually test each hypothesis (causal claims have causal leverage)
• CMB addressed by designed source/time separation, not just a post-hoc test
• Endogeneity strategy specified for HR-adoption / archival causal claims
• Constructs use validated measures; new HR measures piloted; CFA planned
• Intended vs. implemented vs. perceived HR system stated and matched to respondent
• Aggregation justified (r_wg, ICC) with a stated composition model
• Levels aligned across theory, measurement, analysis; nesting modeled
• Power justified for cross-level / interaction effects, not just main effects

Anti-patterns

• Cross-sectional causal claim: "HR practice causes outcome" from one-wave self-report
• CMB as afterthought: a Harman test standing in for designed separation
• Endogeneity ignored: an HPWS→performance coefficient with an obviously self-selected adopter set
• Aggregation by fiat: averaging individuals into a "unit HR system" with no r_wg/ICC
• Construct slippage: measuring perceived practices but theorizing the implemented system
• Underpowered cross-level interaction reported as a null boundary condition

Output format

【Design】multi-source survey / multilevel nested / panel-archival / experiment / qualitative / multi-method
【Hypothesis–design fit】each H testable? notes ...
【CMB plan】source separation + time lag ...
【Endogeneity strategy】(if archival) FE / DiD / IV / natural experiment ...
【Constructs】validated? new (piloted)? CFA? intended/implemented/perceived?
【Aggregation】r_wg / ICC(1)/ICC(2); composition model
【Levels & power】theory/measurement/analysis aligned; power for cross-level/interaction
【Next skill】hrm-data-analysis