Introduction to Bayesian Analysis using Stan

Date:

06/07/2021 - 07/07/2021

Organised by:

Royal Statistical Society

Presenter:

Robert Grant

Level:

Advanced (specialised prior knowledge)

Contact:

training@rss.org.uk

video conference logo

Venue: Online

Description:

This two-day course is ideal for beginners or intermediate users of Bayesian modelling, who want to learn how to use Stan software within R (the material we cover can easily be applied to other Stan interfaces, such as Python or Julia). We will learn about constructing a Bayesian model in a flexible and transparent way, and the benefits of using a probabilistic programming language for this. The language in question, Stan, provides the fastest and most stable algorithms available today for fitting your model to your data. Participants will get lots of hands-on practice with real-life data, and lots of discussion time. We will also look at ways of validating, critiquing and improving your models.

Learning Outcome

  • Use Stan to fit various models to data
  • Check outputs for computational problems, and know what to do to fix them
  • Compare and critique competing models
  • Justify their modelling choices, including prior probability distributions
  • Understand what Stan can and cannot do

 

Topics Covered

  • A quick overview of Bayesian analysis
  • Simulation is useful for statistical inference
  • What is a probabilistic programming language?
  • Parts of a Stan model
  • Univariate models; exploring priors and likelihoods
  • Prior predictive checking
  • Bivariate regression models
  • Predictions and posterior predictive checking
  • Hierarchical models
  • Latent variable models including item-response theory
  • Working with missing and coarse data
  • Gaussian processes
  • Limitations of Stan

 

Target Audience

Anyone with some statistics training who is aware of the advantages of Bayesian modelling could benefit from attending. Fields where this may be most popular are: insurance, political pollsters, finance, marketing, healthcare, education research, psychology, econometrics.

 

Assumed Knowledge

Attendees should be comfortable with using R, Python, Julia or Stata. They should understand probability distributions and basic regression models, though this can be intuitive and doesn’t have to be mathematically rigorous. They do not need to have used Stan before.

Cost:

£588 - £816 (inc. VAT)

Website and registration:

Region:

Greater London

Keywords:

Qualitative Data Handling and Data Analysis, Bayesian Statistics , Bayesian Model , R , Stan

Related publications and presentations:

Qualitative Data Handling and Data Analysis

Back to archive...