<scp>STbayes</scp> : An R package for creating, fitting and understanding Bayesian models of social transmission

Abstract A critical consequence of joining social groups is the possibility of social transmission of information related to novel behaviours or resources. Network‐based diffusion analysis (NBDA) has emerged as a leading frequentist framework for inferring and quantifying social transmission, particularly in non‐human animal populations. NBDA has been extended several times to account for multiple diffusions, multiple networks, individual‐level variables and complex transmission functions. Bayesian versions of NBDA have been proposed before, although these implementations have seen limited usage and have not kept pace with the evolving ecosystem of Bayesian methods. There is not yet a user‐friendly package to implement a Bayesian NBDA. Here, we present a unified framework for performing Bayesian analysis of social transmission using NBDA‐type models, implemented in the widely used Stan programming language. We provide a user‐friendly R package ‘STbayes’ (ST: social transmission) for other researchers to easily use this framework. STbayes accepts user‐formatted data, but can also import data directly from the existing NBDA R package. Based on the data users provide, STbayes automatically generates multi‐network, multi‐diffusion models that allow for covariates that may influence transmission and varying (random) effects. Using simulated data, we demonstrate that this model can accurately differentiate the relative contribution of individual and social learning in the spread of information through networked populations. We illustrate how incorporating upstream uncertainty about the relationships between individuals can improve model fit. Our framework can be used to infer complex transmission rules, and we describe a numerically stable parametrization of frequency‐dependent transmission. Finally, we introduce support for dynamic transmission weights and a ‘high‐resolution’ data mode, which allows users to make use of fine‐scale data collected by contemporary automated tracking methods. These extensions increase the set of contexts that this type of model may be used for.