Bayesian Modeling of COVID-19 Positivity Rate -- the Indiana experience

TL;DR

This paper presents a Bayesian model for estimating COVID-19 positivity rates in Indiana, accounting for variability and overdispersion in daily testing data, providing a straightforward prediction method adaptable by health agencies.

Contribution

It introduces a Bayesian approach specifically designed to model COVID-19 positivity rates with consideration for overdispersion and variability, enabling easy updates and predictions.

Findings

Effective modeling of positivity rate variability

A simple, adaptable prediction procedure

Numerical results via an updatable R Markdown document

Abstract

In this short technical report we model, within the Bayesian framework, the rate of positive tests reported by the the State of Indiana, accounting also for the substantial variability (and overdispeartion) in the daily count of the tests performed. The approach we take, results with a simple procedure for prediction, a posteriori, of this rate of 'positivity' and allows for an easy and a straightforward adaptation by any agency tracking daily results of COVID-19 tests. The numerical results provided herein were obtained via an updatable R Markdown document.