Sample Size and Simulation for Negative Binomial Outcomes

Provides tools for planning and simulating recurrent event trials with overdispersed count endpoints analyzed using negative binomial (or Poisson) rate models. Implements sample size and power calculations for fixed designs with variable accrual, dropout, maximum follow-up, and event gaps, including methods of Zhu and Lakkis (2014) and Friede and Schmidli (2010) as well as extensions for score-test sizing and gaps between events. Supports group sequential monitoring by building on the 'gsDesign' package. Includes recurrent-event simulation utilities (including seasonal rates), interim data truncation, Wald and score-test inference for rate ratios, and information estimation and sample size re-estimation with or without treatment-group labels.

gsDesignNB

gsDesignNB provides design, simulation, and interim monitoring tools for recurrent-event trials analyzed with negative binomial rate models, with Poisson methods available as the limiting special case when dispersion is negligible.

The package is NB-first: plan designs with sample_size_nbinom(), simulate recurrent-event data with nb_sim() or nb_sim_seasonal(), and evaluate group sequential monitoring or sample size re-estimation with sim_gs_nbinom() and sim_ssr_nbinom(). Planning and simulation can use either Wald or score-test inference for rate ratios. The gsDesign package supplies the underlying spending-function and boundary calculations used by those workflows.

Start Here

• sample-size-nbinom for fixed-design planning
• score-vs-wald-simulation for Wald/score sizing and Type I error guidance
• simulation-example and seasonal-simulation for recurrent-event data generation
• ssr-example and ssr-simulation-study for negative binomial SSR workflows

Installation

You can install gsDesignNB from CRAN with:

install.packages("gsDesignNB")

Or install the development version from GitHub with:

remotes::install_github("keaven/gsDesignNB")

Code style

This package follows the tidyverse style guide.