Exercises for Thursday February 26

1. Thanks for modelling efforts for the lifelengths of Roman era Egyptians two thousand years ago. We shall pursue some of these ideas later on in the course.

2. I explained that Pr(Northug wins today) is only 34%, since he reaches the final via the second of the two semifinals. A bit later I will make data for 57 top events (for seasons 2014, 2013, 2012, 2011, 2010) available on the course website, aiming for constructing and selecting among different candidate models, with the purpose of pinpointing the unfairness aspects involved in the current setup for the world championships and Olympics. Vegard Ulvaing is listening.

3. I went through the last bits of Ch 2 and explained the basics behind the BIC of Ch 3. Next week I continue with Ch 3 material.

4. Exercises for next week: Do the birds exercise 20 (a), with 2^3 = 8 candidate models, and supplement this with both AIC^* analysis and cross validation. Compare your results to approximations given by the theory of Ch 2. Then set up a simulation experiment to compute p(delta, n) = Pr(AIC selects wide model) in a setup with n data point from Gamma(1+delta/sqrt(n), theta), with wide model the gamma and narrow model the exponential. Compare your p(delta, n) curve with the limit function given by the theory of Section 2.8. Finally we do Nils Exercise 5.