PSClinical is an interactive R Shiny application for sample size & power computation in clinical trials. It allows users to simulate and visualize empirical power across different study designs, supporting continuous, binary, and survival outcomes under parallel, paired, and one-sample configurations.
This toolkit helps biostatisticians, data scientists, and clinical researchers:
- Explore how design parameters influence statistical power
- Compare analytical vs. simulated power results
- Understand complex time-to-event (survival) scenarios using Weibull or Exponential distributions
| Design Type | Sub-Designs | Key Inputs | Methods Used |
|---|---|---|---|
| Continuous | Parallel, Paired, One-Sample | Δ (mean difference), σ (standard deviation) | t-tests |
| Binary | Parallel, Paired, One-Sample | p₁, p₂ (probabilities) | proportion tests |
| Survival | — | HR, λ₀, accrual, follow-up, Weibull shape | Cox PH simulation |
Empirical power is estimated via repeated Monte Carlo simulations:
Empirical Power = (# of simulations with p-value < α) / (Total simulations)
- Exponential: constant hazard rate over time
- Weibull: flexible hazard, allowing increasing or decreasing risk depending on the shape parameter
You can specify:
- Baseline hazard (λ₀)
- Hazard ratio (HR)
- Shape parameter (k) for Weibull
- Accrual and follow-up durations
- Dropout rate
Users define:
- Design type (Continuous / Binary / Survival)
- Sub-design (Parallel, Paired, One-Sample)
- Sample size, repetitions, significance level, and other parameters
- Distribution and shape for survival data
Displays:
- Analytical sample size or required events for survival
Power estimation for a survival outcome
- Empirical power estimates with Monte Carlo SEs and 95% CIs
- Distribution plots of simulated power values
Power estimation for a parallel design continuous outcome