Beskjeder

A solution set to the exam is available here

Exam information: There will be no permitted aids to the exam (except for the usual calculator). I will however not assume that you remember all the details involved in the course, but focus on general understanding.

A solution-set to the second compulsory assignment is now available from the "compulsory" page.

REMINDER: Wednesday Nov 18 the lecture will be at Norwegian Computing Center, 4th floor, informatics building, room alpha-omega. Afther the lecture by Petter Abrahamsen, there will be a summary of the course, see slides

R-code for extra exercise 1 is now available on the exercise page. This code can be helpfull for the compulsory exercise.

Compulsory assignment 2: There was a misprint in exercise 1 (g) in that a minus sign was missing in the expression for the conditional density. This has now been corrected.

A chapter about Bayesian methods is given as a link to the left (under Examination/form of assessment by a mistake). Those that would like to understand a bit more of such methods could perhaps find this note valuable. It is however not part of the syllabus.

A trial exam is now available. It is much larger than an ordinary (3 hour) exam but illustratew the types of exercises that can be given at the exam. I will not have the capacity to receive and evaluate papers from you, I but will provide a detailed solution set later on that can be used for self-evaluation. Geir

The second compulsory assignement is now available.

Exercise 2007-25 was a bit more difficult than I thought. I have therefore included an extra exercise giving a simpler application of the Gibbs sampler. This exercise should be tried out first. (NOTE: There were some errors in the first version that is corrected now!!!)

Links to a couple of papers about the Metropolis-Hastings algorithm and the Gibbs sampler are put on the "Syllabus" page.

There is a typo in exercise 7.21 (see the exercise webpage). This exercise can be a bit difficult, but we will use some time in discussing it in the lecture.

The first compulsory assignment has now been evaluated. Those not having get it accepted should have got a separate email on this. The assignements can be picked up at "Ekspedisjonskontoret".

Oct 12-14 I will be away, and there will therefore not be any lectures this week. Instead, I will be available for help on the compulsory exercise thursday oct 15 9.30-12.30. Geir

Because of the compulsory assignment, there will not be given any new exercises to oct 7. We will instead use all the three hours for lectures.

First compulsory assignement is now available.

An updated version of the note on MC-essentials is now available (mainly extending it somewhat).

I have now written a small note extracting the main definitions and results needed from chapter 6, see under "Syllabus/achievement requirements". This noe will hopfully be easier to follow than the book on this part.

Some hints to the exercises from 2007 are given on the "exercise" page.

An important part of the course will be to implement algorithms. You are free to use any kind of software. Illustrations during the lectures and exercises will however be based on the statistical package R. Many of the exercises also contain hints on which commands to be used in R. A link to an introduction in R is given under "software".

NBBBBB: The place for the course has been changed to Aud 4 (wedn. 12.15-14.00) and Aud 3 (wedn. 14.15-15.00) in Vilhelm Bjerknes' house .

In the course we will use both some theory about Markov Chains (STK1130) and Bayesian methods (STK4020). We do however not require any of these courses and the main theory that we will need will be introduced through the course. For those wanting a quick introduction to Markov chains, Wikipedia is a possible source. For those wanting a quick introduction to Bayesian methods, see section sec 1.3 in the textbook or Glark and Gelfand, mainly 1.2 and 1.3 but 1.4 may also be of interest.

There will be two compulsory assignments in the course.

Textbook: Robert and Casella (2004): Monte Carlo Statistical Methods, Springer