power-analysis

Sample size calculation for clinical trials requires specifying endpoint types like binary, continuous, or survival data. This Skill computes required sample sizes for superiority and non-inferiority designs while generating interactive power curves and sample size landscapes for visualization.

The Winner's Curse in clinical trials is an overestimation bias of effect sizes common in underpowered studies. This Skill applies shrinkage factors to effect size estimates based on prior evidence quality, mitigating this bias when performing statistical power calculations.

Estimating statistical power for a Phase 3 oncology trial involves applying Winner's Curse correction to Phase 2 hazard ratio estimates. This Skill calculates the optimal sample size needed to achieve target power, such as 80%, to detect clinically meaningful improvements in overall survival.

Non-inferiority sample size calculations for survival endpoints are fully supported. This Skill handles superiority and non-inferiority trial designs across binary, continuous, and survival data, incorporating shrinkage factors to adjust effect size estimates based on evidence source quality.

Visualizing power curves and sample size trade-offs is handled using Python and Plotly for interactive charting. This Skill generates detailed sensitivity analyses, sample size landscapes, and Winner's Curse demonstrations to aid in understanding clinical trial design constraints.

Underpowered clinical trials overestimate treatment effect size due to the Winner's Curse, a statistical bias where only the most extreme positive results achieve significance. Applying shrinkage factors to effect size estimates based on prior evidence quality corrects this overestimation.