Bayesian workflow for disease transmission modeling in Stan

Abstract

This tutorial shows how to build, fit, and criticize disease transmission models in Stan, and should be useful to researchers interested in modeling the severe acute respiratory syndrome coronavirus 2 (SARS‐CoV‐2) pandemic and other infectious diseases in a Bayesian framework. Bayesian modeling provides a principled way to quantify uncertainty and incorporate both data and prior knowledge into the model estimates. Stan is an expressive probabilistic programming language that abstracts the inference and allows users to focus on the modeling. As a result, Stan code is readable and easily extensible, which makes the modeler's work more transparent. Furthermore, Stan's main inference engine, Hamiltonian Monte Carlo sampling, is amiable to diagnostics, which means the user can verify whether the obtained inference is reliable. In this tutorial, we demonstrate how to formulate, fit, and diagnose a compartmental transmission model in Stan, first with a simple susceptible‐infected‐recovered model, then with a more elaborate transmission model used during the SARS‐CoV‐2 pandemic. We also cover advanced topics which can further help practitioners fit sophisticated models; notably, how to use simulations to probe the model and priors, and computational techniques to scale‐up models based on ordinary differential equations.

Dual-Numbers Reverse AD, Efficiently

“…This can be generalised somewhat beyond what we discuss in this paper, but it is unclear how to make the algorithm work in a compositional manner for higher-order primitive operations such as numerical differential equation solvers or root finding routines, which are used in practice for e.g. statistical modelling [Chatzilena et al 2019;Grinsztajn et al 2021],…”

Infection intensity-dependent accuracy of reagent strip for the diagnosis of Schistosoma haematobium and estimation of treatment prevalence thresholds

“…For a 0 a normal distribution with mean -1 and SD 1.5 was chosen, a gamma distribution with shape and scale parameters 5 and 30 for a 1 , a normal distribution with mean 10 and SD 10 for a 2 , and a beta distribution with parameters 10 and 1 for a 3 to ensure a non-informative distribution of sensitivity curves in the relevant range of infection intensities. Priors for the sensitivity of the reagent strip were chosen using prior predictive checks to ensure non-informativity [ 28 ]. Other parameters have semi informative prior distributions over sensible parameter values.…”

Flexible and efficient Bayesian pharmacometrics modeling using Stan and Torsten, Part I

“…Hence the cost of doing posterior predictive checks, even when it involves solving ODEs, is marginal. The computational scaling of Stan, notably for ODE-based models, is discussed in the article by Grinsztajn et al (2021) [33].…”

“…Writing the ODE as a difference from the baseline means the initial PD conditions is 0, as opposed to a parameter dependent value. This results in better computation, because derivatives of the ODE solution with respect to the initial conditions no longer need to be computed; for more details, see [33,Section 5.2]. In addition, we encode a constraint on the circulatory compartment…”