Lecture plan

This page will be updated with a short discription of the content of the lectures. 

Thursday 25 August

  • A general introduction to the course was given, and we went through some of the core material of Chapter 1.1-1.3
  • Lecture notes
  • Simple example

Thursday 1 September

  • We finished the introduction based on key aspects of Chapter 1
  • Lecture notes

Thursday 8 September

Thursday 15 September

Thursday 22 September

Thursday 29 September

  • Non-informative prior distributions
  • More conjugate, multiparameter models
  • Lecture notes

Thursday 6 October

  • The conjugate multinomial-Dirichlet model
  • Started asymptotic theory for Bayesian analysis (ch. 4)
  • Lecture notes
  • The US election poll and asthma mortality rate examples are in this note

Thursday 13 October: Lecture cancelled

 

Thursday 20 October

  • Sketch of proof of the normal approximation
  • Intro hierarchical modelling, start with a real MCMC (Gibbs-)- simulation example
  • Lecture notes
  • The R-script for the Gibbs sampling is here

Thursday 27 October

  • More on constructing hierarchical models (Ch 5.3 of the textbook). Lecture notes
  • Briefly on the empirical Bayes approach (not in the textbook). Lecture notes in this note
  • Peter Müller, who is Professor at The Mathematics Department at the University of Texas at Austin, is visiting us this semester and is going to give a lecture on material from chapter 10 of the textbook. Peter's lecture notes

Thursday 3 November

Thursday 10 November

  • Continue Markov Chain Monte Carlo: 
    • The Metropolis-Hastings algorithm, and special cases (11.2-11.3)
    • Inference and check of convergence (11.4)
  • Intro to Bayesian regression modelling  (approximately 14.1-14.2 and 16.1)
  • Lecture notes
  • R-script for the simple textbook Metropolis example, recreating approximately Fig11.1

Thursday 17 November

  • More about regression models (but the majority of time spent on exercises)
  • Lecture notes

Thursday 24 November

  • More on MCMC, including effective number of independent simulation draws neff (but not the estimation of the autocorrelations, and hence not the estimation of neff)
  • A brief overview of more efficient MCMC algorithms and alternatives to MCMC (not part of the curriculum)
  • Lecture notes 

Thursday 1 December

  • 13.15-max 14.30: Open lecture, for example questions from the students
Published Aug. 23, 2016 2:25 PM - Last modified Nov. 30, 2016 2:55 PM