Isr Methods

Research Design & Method Fit (isr-methods)

When to trigger

• You are unsure which genre best supports your IS claim
• The design may not be able to identify the effect (empirical) or may rest on unjustified assumptions (analytical)
• You are combining methods and need the multimethod logic to hold together
• A reviewer says "the method cannot answer the question" or "the artifact is not evaluated"

Match the genre to the claim

ISR is deliberately pluralistic; no single method is mandated. Choose the genre the claim demands:

Genre-specific design discipline

• Behavioral empirical. Establish construct validity by design (multi-item validated scales, manipulation/attention checks), separate sources/waves to limit common-method bias, justify the sampling frame, and (for experiments) pre-register and power for interactions, not just main effects.
• Analytical modeling. The design is the model: state agents, timing, information structure, and equilibrium concept; defend each assumption; plan the comparative statics and the extensions/robustness that show the result is not an artifact of one assumption. Reserve full proofs for the electronic companion.
• Design science. Specify the artifact, the design objectives, and an evaluation that demonstrates utility (benchmarks, controlled studies, real-world deployment) — a build without evaluation is not a DSR contribution.
• Archival/causal. Name the identification strategy (DiD, IV, RDD, matching) and the threat it addresses; a regression without identification is descriptive.

Sociotechnical level and fit

State the level(s) of analysis and ensure the design observes the level where the mechanism operates (e.g., group-level theory needs group-level variation). Cross-level claims need cross-level data.

Execution bridge (StatsPAI / Stata MCP)

For the empirical / causal lane, estimate and audit rather than only specify. Full map: execution-with-mcp. ISR is empirical IS with strong econometric and experimental work; identification (DiD / IV) for observational claims, randomization inference for experiments.

• 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

• Genre matches the claim; not chosen by habit or data availability
• Empirical: identification/validity strategy named and adequate
• Analytical: assumptions justified; comparative statics and robustness planned
• DSR: artifact + evaluation that demonstrates utility
• Level(s) of analysis observed where the mechanism operates
• Multimethod combinations have an explicit integration logic (ISR 36(2))
• Page budget planned (32-page text cap) with overflow routed to the electronic companion

Anti-patterns

• Method by convenience: using the data you have rather than the design the claim needs.
• Regression theater: archival regressions presented as causal without identification.
• Build-only DSR: an artifact with no rigorous evaluation.
• Multimethod garnish: a second method bolted on without theoretical integration.

Output format

【Claim】[...]
【Genre & design】experiment / qualitative / analytical / DSR / archival / multimethod
【Identification or assumptions】[...]
【Level(s) observed】[...]
【Validity/robustness plan】[...]
【Page/EC budget】[...]
【Next step】isr-data-analysis