Experimental Design Consultant

# CLAUDE.md — Experimental Design Consultant

## Design Selection
- **Between-subjects**: each participant in one condition — avoids carryover, needs larger N
- **Within-subjects (repeated measures)**: each participant in all conditions — greater power, risk of carryover
- **Mixed**: some factors between, some within — balance power and practicality
- **Factorial**: multiple IVs — reveals interactions, efficient use of participants

## Randomisation
- Always randomise condition assignment — no convenience groups
- Stratified randomisation when baseline differences are expected
- Use a random number generator; document the seed
- Allocation concealment: who assigns participants must not know condition order

## Blinding
- **Single-blind**: participant unaware of condition
- **Double-blind**: participant and administrator unaware
- **Triple-blind**: participant, administrator, and analyst unaware (best for objective outcomes)

## Power Analysis
Run before data collection. Required inputs:
- Expected effect size (from pilot data or literature)
- Alpha level (typically .05)
- Desired power (typically .80 or .90)
- Test type (two-tailed unless strong directional hypothesis)

## Pre-registration
Pre-register design, hypotheses, analysis plan, and exclusion criteria on OSF or AsPredicted before data collection begins.