Getting started in Bayesian modelling with STAN and RStan


Rough agenda:

• 5:45: Arrival

• 6:00: Food, drinks & networking

• 6:30: Intro & main presentation

• 7:30-8: More networking

Getting started in Bayesian modelling with STAN and RStan

Bayesian statistical modelling provides a flexible and cohesive framework for parameter estimation, prediction, uncertainty propagation, hypothesis testing, model comparison and evaluation. While Bayesian statistical models have primarily been applied inferentially, they have recently had some successes in prediction competitions such as Kaggle.

Stan ( ( is a relatively new probabilistic programming language that implements novel Markov Chain Monte Carlo algorithms for parameters estimation in Bayesian statistical models. Specifically, the No-U-Turn sampler (NUTS), a variant of Hamiltonian Monte Carlo (HMC), shows improved performance in many situations over the Gibbs / Metropolis-Hastings sampling approaches previously implemented in platforms such as WinBUGS, OpenBUGS and JAGS.

Using the RStan package, which provides an interface between R and Stan, this presentation will provide an introduction to Bayesian modelling, including: installing Stan, formatting data, writing and running basic models in Stan, making predictions and evaluating model outputs.

Speaker Bio:

Bill Dixon is a risk analyst working in the public sector. He holds a PhD in environmental risk analysis and has over 20 years work experience in both government and private sectors. His primary research areas are in environment, health and safety and regulatory compliance with a focus on uncertainty and decision making. He has previously presented to MelbURN on animated mapping in R with ggplot2.


We'd like to thank Talent ( becoming one of our sponsors. They'll be hosting this event and providing the catering.