Skip to content

VABAR/vibass_dev_introductory_training

Repository files navigation

Pedagogical approach of the course Introduction to Bayesian analysis

Objective of the training

What participants will be able to do after the training, that were not able to do before. Avoid verbs like know or understand and focus instead on what this knowledge or understanding enable people to perform.

At the end of the course, participants will be able to start applying Bayesian methods to simple models in their field, having acquired an understanding of Bayesian concepts and becoming aware of the possibilities for more specific and more complex methods for in-depth analysis.

Operational goals

What specific skills (again, in terms of action verbs) are necessary to develop during the training in order to achieve the overall goal above.

  1. Identify the "Bayesian building blocks" (prior, likelihood, posterior, predictive) in any Bayesian analysis and interpret their meaning correctly.

  2. Assess the appropriateness of the basic distributions from the exponential family for the various types of data (count, binary, continuous).

  3. Summarise inferences and predictions of quantities of interest and derived functions using random samples from posterior probability distributions.

  4. Interpret the information contained in a posterior probability distribution by visual inspection of its shape in a plot of the density or mass function.

  5. Synthesise the main characteristics of three core approaches for Bayesian computation: analytical conjugate derivations, Importance Sampling and Markov Chain Monte Carlo.

  6. Apply basic Importance Sampling and MCMC methods via available software for fitting regression models

  7. Assess the convergence quality of the Markov Chains by visual inspection of their traces and by convergence indices like $\hat R$.

  8. Summarise the landscape of available software packages, and their characteristics.

    Specifically, whether they are flexible probabilistic programming languages (Bugs, Stan, Nimble, ...) or interfaces to a closed set of families of statistical models, and their relative strengths and weaknesses.

  9. Identify in which situations Bayesian analysis is particularly advantageous with respect to frequentist alternatives.

    Specifically, the direct probabilistic interpretation of results, the treatment of the uncertainty in the model parameters (e.g. proper propagation, rather than marginalisation using MLEs), not relying on asymptotic theory, the probabilistic approach to hypothesis testing.

Pre-requisites

  • Target population: people with little to no previous knowledge of Bayesian statistics with some concrete motivation.

    E.g. master/PhD students who are going to use or work with Bayesian statistics, professionals in various fields (medicine, biology, ecology, industry, etc.) wanting to improve their quantitative skills.

Releases

No releases published

Packages

No packages published