Messages

Exams take place in *Rom 1020*, 10th floor, Fri 10th 9-13 and Mon 13th 9-13. Each candidate needs to check how to work the equipment. Think *40 minutes, max 45*, then with questions and comments and discussion, with sensorene Nils + Emil.

Time table:
 
Fri Dec 10:
9:15 Dennis Christensen
10:15 Lars Olsen
11:15 Per August Moen
12:15 Fabian Bull

Mon Dec 13:
9:15 Ingrid D?hlen
10:15 Adam Rogers
11:15 Haifeng Xu
12:15 Haris Fawad

* Dennis Christensen: f?r K. Doksum, 1974, Annals of Probability, "Tailfree and neutral Bayesian random probabilities and their posterior distributions". (Kjell Doksum was born in July 1940, and died last week, November 2021.)

* Lars Olsen: f?r Cynthia Rudin, 2019, "Stop explaining black box machine learning models for high stakes decisions and use interpretable models instead".

* Per August Moen: f?r C. Cunen, G.H. Hermansen, N.L. Hjort, JS...

Just for the record of these board messages, we have been and are still following the gameplan4 schedule of things. So on Tue Oct 26 we had

* Ingrid D: Efron, annals 1975

* Per August M: Angrist, Imbens, Rubin, jasa 1996.

On Tue Nov 2 we have

* Adam R: Montoya 2019

* Fabian B: Welling and Teh 2011.

On this Tue Nov 2 Nils gives a talk on kortstokkproblemet (and are the millions of decimals of pi really iid uniform on 0-9?), just after the stk9200.

For Tue Oct 19: Lars O presens Efron (1979), the birth of bootstrapping; Dennis C presents Ferguson (1973), the birth of Bayesian nonparametrics.

For Tue Oct 26: Ingrid D presents Efron (1975), on two ways of looking at and using logistic regression; Per August M presents the Nobel Prize Class paper Angrist, Imbens, Rubin (1996).

I've uploaded more papers on the course site, and also updated the gameplan3 document.

I will have a "gampland3.pdf" at the website soon, with more details for the coming weeks. At any rate, for Tue Oct 12, we have: 

* Haris F: presents Ryalen, Stensrud, R?ysland, their LIDA paper 2019. 

* Dennis C: presents Ferguson, his Annals of Statistics paper 1973. This paper is in both categories A and B. It's the famous start of Bayesian Nonparametrics. Check also Ferguson's 1974 paper, also on the reading list on our website, though we focus on the 1973 Dirichlet processes paper. 

1. Thanks to Haifeng X and Fredrik W for giving inspiring presentations of Santoro et al. (2016), with machine learning, and Tibshirani (1996), the first Lasso paper. 

2. For Tue Oct 5, we have

* Thomas M, on Ruczinski et al. (2003), logic regression; 

* Kristina B.H., on Ribeiro et al. (2016), why should I trust you? 

1. Inspiring & ambitious presentations, Tue Sep 21, by Haris F on Holland (1986) and by Lars O on Aas, Jullum, L?land (2021); the final version of the latter paper is now on our website. I'm also uploading more papers and presentations.

2. For Tue Sep 28, we have Haifeng X on Santoro et al. (2016), with some Google DeepMind people on board, and Fredrik W on Tibshirani (1996), the famous first Lasso paper. 

3. More information next week, also regarding exam dates.

Thanks to Ingrid D and Fabian F for entertaining presentations of Mikosch (2006, with discussion) and Hastings (1970), Tue Sep 14. We had time for various pendants and tangents and reflections and predictions.

For Tue Sep 21, Haris F presents Holland (1986), on causality, and Lars O presents the recent Aas, Jullum, L?land (2021), on Shapley values.

We might use five minutes, before Haris and Lars, to have additional comments and questions regarding the previous presentations and questions raised there. Again, I recommend the MCMC exercise I came up with: find 4 x 4 magic squares, with numbers 1 to 16, such that each row, each column, both diagonals, have sum 34 -- by setting up a probability model and letting MCMC give you 4 x 4 squares from that distribution, via Metropolis or Metropollis-Hastings.

I've uploaded more papers on the Readling List, and will also post an updated gamplan2.pdf, with further comments. Check also our "hjelpetr?d C", by direct mail.

Another good meeting today, Tue Sep 7, with (1) Adam R giving Hayes and Rockwood (2017) nad (2) Per August giving Fearnhead (2006). We discussed back & forth, and might also use 5 minutes at the start of next Tue Sep 14 to discuss further.

I'm placing a "gamplan1.pdf" on the course site, with tentative time plan from the start of the Journal Papers Reading Club Course to mid October.

Check the "hjelpetr?d C" mail I've sent to you for more details; there'll be further hjelpetr?der in weeks to come. You may also give your comments or suggestions there.

For Tue Sep 14, we need to be strict with the time slot, as I need to leave the room at 13:47 (to be "virtually present" in Bruxelles from 13:57).

First we use 5 minutes for generalities. Then time slot approx 11:20 to 12:20: Ingrid D gives Mikosch (2006), who cites H.C. Andersen, and manages to make at least some discussants skikkelig irriterte. Time slot approx 12:35 to 13:35:...

On Tue Aug 31, Dennis Christensen gave a wondrous presentation of Breiman (2001), and the class behaved vigilantly and vigorously. We used more than two hours! 

His pdf, and those from later presentations, will be put on the course website. Tentative ambition for us: each candidate gives two presentations, during the semester. 

I'm placing more papers on the Reading List as we go along.

For Tue Sep 7, we do *two* papers. (1) Adam Rogers gives Hayes and Rockwood (2017, the BRaT journal), so we'll learn about moderation analysis. (2) Per August Moen gives Fearnhead (2006, Stat Computing), with Bayesian inference and algorithms for multiple changepoints. 

Thanks for all your diligent homework. I'll produce a note, to be group-mailed to all of you, with tentative plans for the coming few weeks. 

On Tue Aug 24, we had a good introduction meeting for the course. We'll need a couple of weeks to make further decisions, particularly when it comes to the Reading List itself. But Paper 1 for us is Leo Breiman's "Statistical Modeling: The Two Cultures", Statistical Science, 2001 (with discussion contributions).

We all read it!, and it will be presented by Dennis Christensen, Tue Aug 31. Take notes, check your own views, and take active part in the discussion. 

*Homework* for all of you: Within Mon Aug 30, send me a mail, with (a) some sentences about your PhD project, and (b) suggestion of *two papers*, from the relevant research area, that could land on our Reading List. 

Welcome to the course -- which is a different type of course than the usual ones. It consists in *reading a selection of journal papers*, where the students, or is it candidates, present them to the class, with discussion and supplementing information, etc. 

The time slot given us for the course is Tuesdays, 11:15 to 14:00, in Undervisningsrom 1119, but we will not use the full 3 hours every week; this depends also on the number of candidates. 

We'll use the first couple of weeks, from the course starts up Tue Aug 24, to set up such a list of papers to be read, and a bit later on these are assigned to the different students. It's not meant to be a "free ride" for the others, as all students are meant to work through all papers. For the exam, student X, who has presented papers A, B, C in the course, might be asked to present papers D, E, so to speak.

By definition we come back to various practicalities later. Have a look at t...