CRTpowerdist

CRTpowerdist performs power analysis for cluster randomized trials (CRTs) with unequal cluster sizes by calculating attained power conditional on realized allocations and constructing pre-randomization power distributions to assess risk of inadequate statistical power.

Key Features:

• Attained Power Calculation: Calculates attained power conditional on realized allocations post-randomization to quantify actual statistical power for a given allocation.
• Power Distribution Construction: Constructs a pre-randomization power distribution based on potential allocations generated by the randomization algorithm (RA) to evaluate risk before trial commencement.
• Support for Trial Designs and Outcome Types: Supports cross-sectional stepped-wedge and parallel cluster randomized trial designs, with or without stratification, for continuous (Gaussian), binary (Binomial), and count (Poisson) outcomes.
• Analytic Formulae and Simulations: Implements simulation-based approaches alongside approximate analytic formulae to estimate attained powers across realized and potential allocations.

Scientific Applications:

• Risk Assessment: Evaluating the likelihood of inadequate attained power due to randomization variability in CRTs.
• Algorithm Evaluation: Assessing and comparing randomization algorithms (RA) by their propensity to produce unfavorable allocations affecting power.
• Design Optimization: Informing trial design decisions, including stratification and allocation schemes, based on distributions of attained power.

Methodology:

The package computes attained power conditional on realized allocations, constructs pre-randomization power distributions from potential allocations generated by the randomization algorithm (RA), and uses simulation-based approaches together with approximate analytic formulae to estimate power for the supported CRT designs and outcome types.