Messages

We have covered a decent amount of themes and topics during the course. Our last teaching encounter takes place Mon May 30, where I will (1) put up the essential CDs for the simplest of meta-analysis setups, from Ch 13, and (2) look back at the more important issues of the course. My presence is required at some Abel festivities with Sir Andrew on Mon May 23, so there's no teaching then (I'll also notify some of you by ordinary mail about this).

Please contact me if there are particular themes or issues or details (theoretical or "practical") you might wish to have discussed when we meet on May 30.

As previously agreed, the *exam project exercise set* will be made available at this course site on Wed June 1, with reports to be delivered, in duplicate, by Mon June 13. The four-hour written exam is on Thu Jun 16.

As was mentioned during the lectures, there's no teaching in the course on Mon May 9, but those who wish to are welcome to attend some of the lectures at the FocuStat FICology Workshop at Teknologihuset, Mon-Tue-Wed May 9-10-11. See the FocuStat webpage for the programme.

http://www.mn.uio.no/math/english/research/projects/focustat/news-and-events/

For Mon April 25 I plan two hours of exercises and one hour lecture. For the exercises, we start out with these from Chapter 8: 1, 2, 3, 4, 14. We then do those mentioned earlier from Chapter 7.

Greetings from Milano, where I'm giving a somewhat intense short chourse on the CLP.

I take part in the BFF conference (Bayesian, Frequentist, Fidicual; Best Friends Forever) at Rutgers April 11-13, incidentally giving a lecture based on CLP material. There is therefore no teaching Mon April 11.

For Mon April 18, work through exercises 7.1, 7.3, 7.5, 7.9. Concentrate on becoming familiar with two of the central methods of Chapter 7: t-bootstrapping and Bartletting the deviance function. Check out how these two methods work when n datapoints are observed from the normal (\mu,\sigma^2), where the focus parameter is \psi = \mu + 0.675 \sigma, the upper quartile.

I'm uploading R scripts com17a (birds on island) and com18a (cc for gamma parameter a) to the site. Check these, and try extensions and alternatives, e.g. when it comes to CD for the \tau parameter for the birds.

There's no teaching in the Easter Week, and this includes Easter Monday, 28 March, which means that our next CLP encounter is Mon 4 April. I then continue with Ch 5 material. Ch 6 is not part of the curriculum, and for Ch 7 I will focus on Sections 7.3, 7.4, 7.6.

For exercises: A. Work through the Neyman-Scott cousin problem, where n pairs (Y_{i,1},Y_{i,2}) are observed with the same mean (\mu,\mu), but with different \sigma_i. Work out the log-profile likelihood function for \mu and check what the Wilks theorem recipe leads to. Then repair this to do better. B. Exercises 4.17, 4.18. C. Work through the details of Example 5.10, with both the optimal and approximate CDs, for a dataset of size n = 5, as given in Exercise 7.9.

On Mon 7 March I went through more of Chapter 4 material, including inverse regression, multimodal likelilhoods and briefly about survival analysis models. Next week we start Chapter 5, where about half the material will be seen as "cursory curriculum". For exercises, I've uploaded com13b which does cc(p) for logistic regression, for given covariates.

Exercises for Mon 14 March: 4.3, 4.4, 4.12, 4.19.

The CLP book is now in Akademika! You ought to buy it, there or perhaps via amazon.

On the skuddårsdag I went through some further parts of Ch 3, including CDs for change-points, and started on Ch 4. We discussed "lazy statistician's confidence curve", where the log-likelihood profile and hence deviance is computed by a lazy unconstrained operation as opposed to a clear constrained optimisation; see the R script com9a for the 2 x 2 table. We also went through the Lagrange mulitiplier method.

Exercises for Mon 9 March: 3.12 (with dataset on 189 mothers from the Claeskens and Hjort book website), 3.14, 3.16, 4.3, 4.6.

On Mon 22 Feb I discussed various topics from Ch 3, including that chapter's main recipes and tools for finding CDs and ccs. I also did Exercise 3.1. For Mon 29 Feb I will start Ch 4, and spend about half of the time for exercises, namely 3.4, 3.5, 3.9, 3.11, plus the following. 44 women and 101 men have been asked about whether they are positive to same-sex marriages or not, for the US state Vermont. For the women, 35 said "yes", 9 "no". For the men, 60 said "yes", 41 "no". Treat the data as a multinomial 2 x 2 table, with underlying probabilities p_{i,j}. Test independence by giving a full confidence curve for the appropriate \gamma = \sum_{i,j} (p_{i,j} - a_i b_j)^2/p_{i,j}, where a_i = Pr(X = i) and b_j = P(Y = j), with X = gender and Y = yes/no to the question. Also find one or two relevant focus parameters \psi and give confidence curves for these.

On Mon 15 Feb, Céline Cunen had her Leonard Bernstein moment, with Nils away at the Olympics. She went through several exercises and elements of Sections 3.3, 3.4. On Mon 22 Feb, I will go on with Ch 3 topics, also touching situasions with discrete parameters.

For exercises, work with 3.1, 3.4, 3.5, 3.9, 3.11.

On Feb 8 we rounded off Ch 2 and started on Ch 3. I also discussed various exericse.

I've now uploaded *a couple of R scripts* and also an updated version of the Nils Exercises.

For Mon Feb 15, do Nils Exercise 4, then Exercises 2.5, 2.6, 2.7 from CLP.

On 1 February I went through the most important aspects of Ch 2, including log-likelihood, maximum likelihood, profiled log-likelihood, deviance, with the central results concerning (a) the ML estimator itself and (b) the Wilks theorem for the deviance. I also discussed pivots piv(\psi,\data) and their role in constructing CDs. Time was also used to go through various aspects of Nils Exercises 1 and 2. 

For Mon 8 February, prepare all of Nils Exercise 3, along with Exercises 2.2, 2.3 from the CLP. I will also start on Ch 3.

Within a day or two I'll place R scripts on the course site, for handling details of Nils Exercises 2 and 3.

Mon 25 Jan I went through aspects of Chapters 1 and 2, with discussion of the two most theorems on maximum likelihood behaviour. Next week we continue Chapter 2, partly starting on Chapter 3.

Note that I've now placed "Lecture Notes & Exercises" on the course website. For next week, work through Exercises 1, 2, 3 of these.