Chapter 1 Introduction

PoissonERM is a tool that was developed as an automation tool for Poisson Regression of binary events (eg occurrence of AEs) for internal use at Pfizer Inc. It implements a full statistical analysis using Poisson Regression on endpoint(s) specified by the user. There is also the ability to predict the incidence rate using user provided simulated exposures (usually from a separate population PK model). PoissonERM is the package version of this tool, which can generate an R Markdown report summary that includes all the components typically included in a formal report.

Poisson regression of binary outcome data utilizes the features of the exponential family of distributions. For the purposes of the analysis that is performed with this package, there are two aspect to highlight.

  • the link function for Poisson regression is log() instead of logit;
  • Poisson regression can incorporate time to model the incidence rate.

This tutorial is a document of the methods implemented in PoissonERM and the instruction for users.

Section 2 Quick Start With the Example is for users to start with provided example folders (https://github.com/yuchenw2015/PoissonERM-Example). It shows how to run functions in PoissonERM when the control scripts and data sets are ready. Section 7 provides ideas for users to explore the different options in the example control script.

Section 3 Usage introduces the basic uses of 3 functions in PoissonERM.

Section 4 Data Preparation shows the data structure of the modeling data set and of the new simulated exposure data set.

Section 5 and Section 6 are detailed explanations of the two control scripts.

Ideally, user should prepare the data sets and the control scripts before running functions in PoissonERM.

To install the package, go to https://github.com/yuchenw2015/PoissonERM.

The full section of the analysis method can be found here: https://yuchenw2015.github.io/PoissonERM.

The examples mentioned in this tutorial are here: https://github.com/yuchenw2015/PoissonERM-Example.