Messages

Thanks for valiant efforts, 9 + 3 = 12 of you, with The Oblig. Check the response comments I've posted for each of you, and further details and comments in class. 

We round off Ch5 Mon Nov 9: do 5.30 ("we can do things"), read up on profiled log-likelihoods and Wilks. 

Tue Nov 10: Siem does his LEFSA, 20 minutes, and Pr(some exam detail will flow from this) > 0. Tobias does the beta-binomial thing for the overdispersed children of Story #2. 

The plan is then 3 hours for Ch7, 3 hours for Ch8, 3 hours for Ch9, and then more stories, substories, earlier exam sets. Start out with Exam 2024, #1: 

For Ch7: 1, 2, 3, 5, 8, 19 (and more, inside some stories). 

https://www.mn.uio.no/math/english/research/projects/focustat/lecture-notes-and-exercises-for-various-courses-gi/exam_stk4011_2024.pdf

Nils Lid Hjort holder foredrag ved Fagligpedagogisk Dag, torsdag 29.10, Aud. 1, Vilhelm Bjerknes, kl 14:15 til 15:00 -- kanskje kommer det noe stk4011 i noen fotnoter mot slutten.

** Hvordan løse sudoku og finne magiske kvadrater ved hjelp av sannsynlighetsmodeller og Markovkjeder.

Bruk tallene 1 til 36 til å lage et magisk 6 x 6 kvadrat, slik at hver av de 6 linjer, hver av de 6 kolonner, og (helst også) begge diagonalene, har presis samme sum. I foredraget ser vi på hvordan man kan finne frem til slike magiske kvadrater, ikke ved å være smart tallteoretiker, men ved (a) å lage en smart nok sannsynlighetsmodell over alle mulige 6 x 6 kvadrater, som favoriserer gode forsøk, og så (b) lage en Markovkjede som simulerer fra denne. Etter en million simuleringer finner vi noen slike magiske kvadrater. Metodikken fungerer også til å løse sudoku og andre magiske oppgaver.

The OBLIG is to be delivered Mon Oct 26 (and the writer of these sentences looks forward to these reports). 

We're round off Ch4, and will be starting Ch5, with this subset of execises defined as curriculum: 1, 2, 3, 6, 7, 8, 10, 13, 14, 15, 22, 24, 29, 30, 38, 42, 44, 46. Gently asterisked ones: 7, 8, 10, 14, 22, 29, 30 ("we can do things"), 42.

Stories: #1 cooling of newborns; #2 overdispersed children; soon more.

Check the curriculum list, which is being updated regularly. After Ch5, there will be relatively smaller parts of Chs 7, 8, 9.

https://www.uio.no/studier/emner/matnat/math/STK4011/h26/oblig_stk4011_2026.pdf 

Here is the pdf -- also to be found near the bottom of the course page. Note this, on "page 0", accompanying other words of instruction: 

 Importantly, I ask you to write up half a page or so to tell us how the work proceeded: what was easy; what was not so easy; what was perhaps in the fun category; if you were stuck, how did you unstuck yourself; how satisfied are you with your work. And, since this is 2026 A.D., in the middle of the Manhattan Project of Our Time: did you use AI, and, in that case, how.  

This can suitably be placed at the end of your report (or, for that matter, on your report's opening page, depending on your literary instincts). 

You will need the " twoworlds-data " file, placed under "some datasets used".

Enjoy [imp...

Thanks to Andreas who spotted the mistake! Here is (hopefully!) a correction. /studier/emner/matnat/math/STK4011/h26/ex3_17.pdf

Check the steadily updated curriculum list. For the week starting Mon Sep 28, we round off Ch 3, see list below, with some emphasis on 8, 12, 15, 27, 29. If Nils is in Risør Tue Sep 29, then Emil takes charge that day.  

From Chapter 3: Exercises 1, 2, 3, 4, 5, 6, 7, 8, 9, 11, 12, 15, 16, 17, 20, 25, 26, 27, 29, 32, 33, 37. Gently asterisked: 11, 16, 25, 32, 33. Struggle through, among those mentioned: 8, 12, 15, 27, 29.

From Chapter 4: Exercises 1, 2, 4, 5, 8, 10, 13, 15, 18, 19, 21, 24, 25, 26, 27, 28, 30, 33, 43.

https://www.youtube.com/watch?v=vurLuoDs4ic 

" C. R. RAO - The Man Who Transformed Statistics! " Ok, consider it inside the curriculum. Rao (1920 - 2023) is a giant -- but I'm not entirely sure I like the AI-made aspect of this otherwise nice & informative 33 minute video. Exam questions in December 2026 will touch some of the issues discussed here.

We plod along, with more exercises from Ch 2, from the list 1, 2, 6, 7, 8, 12, 14, 17, 21, 29, 31, 34, 35, 45, 47, 48, 49, 50, 60; gently asterisked: 6, 11, 12, 14, 21, 29, 31, 35, 49, 50. Then from Ch 3, see the updated curriculum list: 1, 2, 3, 4, 5, 6, 7, 8, 9, 11, 12, 15, 16, 17, 20, 25, 26, 27, 29, 32, 33, 37; gently asterisked: 11, 16, 25, 32, 33. 

OBLIG dates: t_0 = Mon Oct 12 to t_1 = Mon Oct 26. 

There will be three (or four?) 20-minute presentations by our PhDs, inside the usual lecture hours; time slots & more to be decided on later.

We've started Ch2 -- check the subset of exercises on the curriculum-to-be page: Exercises 1, 2, 6, 7, 8, 12, 14, 17, 21, 29, 31, 34, 35, 45, 47, 48, 49, 50, 60. Gently asterisked: 6, 11, 12, 14, 21, 29, 31, 35, 49, 50. 

Torjus did 1.45; Siem did 1.54; Tobias does 2.8; more to follow on the blackboard.

Mon Sep 7: we do the Story #78, Poisson processes & waiting times. Soon: Story #83, checking out the CLT. Note: in his real-time-ranting following the Markov inequality and things frå dei nærskylde umråde, Nils did a mistake in the "let's see what Markov says about Pr(|N(0,1)| \ge 1.96)" -- he blames the students for not immediately spotting this. So do it, now, and check more. If Norwegian men's heights are N(1.80,0.09^2), show from Markov that at most 4% are taller than 2.25 ... and comment on this.

Check the "tentative curriculum list", found here on the course site, which will be adjusted & updated a few times during the course. It will list the relevant subsets of chapters 1, 2, 3, 4, 5, 7, 8, 9, and also the list of Statistical Stories we work through. 

In particular, these are for Ch2, for the two-three weeks starting Mon Aug 31:

1, 2, 6, 7, 8, 12, 14, 17, 21, 29, 31, 34, 35, 45, 47, 48, 49, 50, 60. Gently asterisked: 6, 11, 12, 14, 21, 29, 31, 35, 49, 50.

Mon Aug 31 we do Stories #77, #78, and Torjus does 1.45.

During the course we'll be working through perhaps a dozen stories (or subsets of points for these stories). Here are the two first, for the oncoming few weeks, from Mon Aug 31 onwards: 

.. Story #77, handball. Files: (i) "denmark-norway-nov2022-data", 60 action-filled minutes, goal by goal (in minutes + seconds), who wins? (ii) "handball_data_4", 115 match results. (a) Do Poisson things and attempt to do Fig 19.12, left. (b) Also look into the "positively correlated handball processes", finding formulae for Pr(X=x,Y=y) with X = X_0 + R, Y = Y_0 + R, attempt some plots to find a good \rho. 

.. Story #78, football, Belgium coming back from 0-2 and then winning 3-2 over Japan. This is Poisson with exponential times calculus.

In a bit of time we'll be doing bits of Story #2, so have a look, with binomial overdispersion and the beta-binomial model for the 38495 Sachsen 1899 families with 8 or more children. Is p = Pr(girl) constant; no, it's not; etc.

The course material will essentially be (a) certain portions from Chs. 1, 2, 3, 4, 5, 7, 8, 9 (with prior estimates for their proportions being 33%, 33%, 66%, 66%, 50%, 10%, 10%, 10%) and (b) a selection of up to perhaps ten of the 100 Statistical Stories.

Hjort-Stoltenberg 16aug2026 files have now been placed here (and some earlier versions been taken away). The tentative programme for the first few weeks will be these, from Ch1 (with numbers, here and later, being those used in the 16aug2026 files):

1, 2, 3, 6, 7, 8, 12, 13, 14, 16, 
18, 22, 24, 25, 32, 33, 42, 43, 45, 54.

We might be using 2.5 weeks to get us more or less through this, and after that we jump into the core material of Chs. 3, 4. And soon enough we'll be doing the first couple of Statistical Stories.

In the previous message, from June 16, I describe & point to the teaching material, preliminary versions of a subet of the chapters, and lots of Statistical Stories, from the to-be-completed Hjort & Stoltenberg book -- and I uploaded various 3July2026-files. 

There will be new and updated versions of these, where I aim at having 15Aug2026 files ready on that date -- and in that process I do take away the 3July2026 files. 

Nils Lid Hjort 

Welcome to Statistical Inference Theory! The course material will be based on subsets of preliminary chapter versions of the book Statistical Inference: 600 Exercises and 100 Stories (Nils Lid Hjort and Emil Aas Stoltenberg, to be published later in 2026). These preliminary chapters will be published here, around July 1. 

More specifically, the course curriculum will essentially consist of chunks-but-not-all of Chs 1, 2, 3, 4, 5, 7, 8, along with a suitable selection of 12-15 of the book's 100 Statistical Stories.

You may also check the course website for 2025 and 2024, with their Obligs, exam sets, stories, and more. 

Nils Lid Hjort