Bayesian statistics with R

Course level: Intermediate

Next live dates: TBC (Contact us to request training details).

Course description

Throughout our lives, in everything we do, we will have some prior belief about what the result of an action will be, often based on previous experiences. For example, when I take the train to London, I expect to arrive around 20 minutes later than scheduled as UK trains are notoriously unreliable! Bayesian inference allows us to take account of our prior beliefs and incorporate them into our analysis using Bayes’ theorem. In the past, Bayesian analysis has not been as widely used as frequentist, or ‘classical’ methods, as they typically require more computationally intensive methods to carry out. However, vast improvements in the efficiency and accessibility of statistical software has eroded these barriers, making Bayesian statistics more popular than ever.

This online course will introduce Bayesian ‘philosophy’ and show how Bayesian methods differ from frequentist (or ‘classical’) statistical approaches. The course blends theory and hands-on experience to ensure participants understand Bayesian concepts, such as prior and posterior distributions, can apply Bayesian methods to analyse data and interpret their results in a Bayesian manner, and can check these results are valid and robust, all using R software.

The course is designed to be highly interactive with a focus on practical applications, ensuring that you can immediately apply what you learn to your own data. Throughout the course, we will discuss best practices for reproducible coding.

Outline

Topics covered in this course include:

• Introduction to Bayesian thinking: how is it different to classical statistics and why bother?
• Bayes’ theorem and its components: prior, likelihood, and posterior distributions.
• Bayesian statistics in R: a comparison of linear model fitting with classical vs. Bayesian methods.
• Bayesian approaches to model fitting, their benefits and drawbacks.
• Bayesian model comparisons and diagnostics.
• Using Bayesian methods to fit models of differing complexity.

Target audience

This course is designed for anyone curious about Bayesian statistics, or who would like to incorporate their prior beliefs into analysis.

Participants do not need prior knowledge or experience of Bayesian statistics to take this course. Although a recap is included in the course, participants are expected to be comfortable with the definitions and interpretations of p-values and confidence intervals ( see here for a free introduction to inferential statistics). Participants are also expected to have some prior experience of using R to load, tidy, and visualise data, preferably using Tidyverse packages.

Learning outcomes

By the end of this course, participants will have a solid understanding of Bayesian principles, and can distinguish between Bayesian and frequentist ways of thinking. They will be able to confidently apply Bayesian analysis to data, including selecting an appropriate prior distribution, fitting a model using a Bayesian approach, and interpreting the posterior distribution, all using R software. Additionally, they will learn how to diagnose and validate their analyses, ensuring reliability and accuracy of their results.