experimental-design

What problem does it solve?

Experimental research often fails when treatment assignment, sample size, or validity threats are poorly specified, leading to biased estimates and underpowered results.

Core Features & Use Cases

• Experimental & quasi-experimental design guidance: Choose between lab, field, survey, natural experiments, and A/B testing while accounting for internal/external validity tradeoffs.
• Power analysis and MDE planning: Compute required sample sizes, minimum detectable effects, and cluster-randomized design effects to avoid underpowered studies.
• Randomization, balance diagnostics, and validity threats: Implement simple/stratified/cluster randomization and assess balance on observables; address attrition, noncompliance, SUTVA/spillovers, multiple testing, and common reporting pitfalls.
• Analysis framework and pre-registration structure: Use ITT as a primary approach, plan robustness/heterogeneity checks, and prepare a pre-registration outline for repositories like OSF.

Quick Start

Use the experimental-design skill to produce a full RCT or field experiment plan with power analysis, a randomization procedure, balance diagnostics, and a pre-registration-ready analysis outline.