Model Averaging

A comprehensive overview of model averaging, including recent developments, based on the book "Model Averaging" by David Fletcher (Springer 2018)

“This book provides extensive coverage of the literature on model averaging, and an exceptionally clear exposition of the key ideas, illuminated by well chosen worked examples. It is concise and to the point while offering sound advice on practicalities.

Statistics could do with more books of this sort.”

Professor Simon Wood, University of Edinburgh

Outline of the Course

Model uncertainty arises when no single model clearly fits the data better than the other models. Model averaging is a means of estimating parameters

in a way that allows for this extra level of uncertainty. It provides better point estimates and confidence intervals than obtained using a single best model.

In this course you will learn how to make best use of modern methods for model averaging, from both a Bayesian and frequentist perspective. It is based on, and expands upon, David Fletcher's book (Springer 2018).

There is a mix of recorded video lectures and exercises using R, plus a review of R packages for performing model averaging.

It is expected that you would need around 30 hours to complete the course. We are planning to offer live discussion sessions with David Fletcher later in the year, and participants of this course will be able to access these at a discount rate.

Course Teacher:

David Fletcher

David has 35 years of experience in statistical research, teaching and consulting, and has published over 80 scientific papers in international journals.

He is well-known for his papers and book on model averaging, as well as his papers on overdispersion, mark-recapture models, zero-inflated data, and population modelling.

He has provided statistical advice and run workshops for a wide range of scientists and organisations for over 30 years, and enjoys the combination of theory and application in consulting work.

You can find more details about David on his website.