In Bayesian modelling, we use Bayesian inference as a procedure to estimate parameters of interest. This procedure allows incorporating available knowledge about the said parameters as “priors”, that combined with observed data as “likelihood” provides us with updated estimates known as “posteriors”. Through these components, one can construct custom models and specify probabilistic conjectures about natural phenomena in terms of priors and likelihood. As a direct advantage to the inference procedure, one gets access to an entire posterior distribution of the parameters of interest that retains transparency in terms of treating uncertainty in estimation. Bayesian models can be easily constructed and implemented using R and Stan. Stan is a C++ library for Bayesian modelling and inference that primarily uses HMC: Hamiltonian Monte Carlo (specifically the No-U-Turn sampler, or NUTS developed by Hoffman and Gelman [2012]), to obtain the posterior samples given a user-specified model and data. The R package rstan provides Rstan, the R interface to Stan. The rstan package allows users to conveniently fit custom Stan models from R and provides access to outputs in terms of posterior samples as well as several downstream model checks, both numerically and visually. In this talk, I will discuss about the advantages of Bayesian modelling, how to build custom models in Stan and fit them using R, methods of model checking as well as briefly introduce other advanced R packages that allow easy access to Bayesian modelling and inference.
