{
  "study_id": "NCT07611422",
  "agent": "baseline",
  "run": 3,
  "p_primary_hypothesis_supported": 0.18,
  "effect_direction": "favours intervention",
  "effect_size_90pct_interval": {"low": -0.2, "high": 0.45, "metric": "Cohen's d for positive-information vs neutral-information on PSS-10 at 4 weeks (positive = less stress with positive information)"},
  "moderator_prediction": "Expectancy effect slightly larger in the cold-shower arm, where a plausible, widely publicised stress-resilience rationale and a salient daily ritual reinforce the suggestion; the warm-shower rationale is less credible. Interaction unlikely to reach significance at n=120.",
  "claim_ids_used": [],
  "rationale": "Single verbal expectancy manipulation, healthy volunteers, outcome is PSS-10 (one-month recall trait-like stress) four weeks later. Primer: expectation effects are large for acute self-report but small for chronic/delayed outcomes; OLP in non-clinical samples ~0.29 and likely smaller for a one-off rationale. Realistic d ~0.1-0.2. With ~60 per expectancy level, power to detect d=0.2 is ~20%. Snowball sample, low baseline stress limits room for change. Participants are single-masked but the rationale is salient, favouring small self-report shifts. Alternate path via expectancy x time interaction adds a little. Pre-registered psychology base rate ~44% adjusted down for power. Estimate ~0.18."
}