News and Events
Here we list various news & events related to the FocuStat project and the people involved, also concerning our own workshops and research kitchens. Check also our Facebook page for additional information, discussion, and some photos.
War-and-Peace Changepoint Analysis Points to the Vietnam War 1965 – and Pinker likes it
The PRIO event on January 11, with Nils Lid Hjort, Håvard Hegre and Siri Aas Rustad in the panel, discussing a lecture by Aaron Clauset, caused Nils to let go of a skiing Saturday, to write down his Towards a More Peaceful World [insert `!' or `?' here] FocuStat blog post. The post reports on various statistical modelling and interpretation issues for a Correlates of War dataset on the onset and body counts of serious wars, across the 1823 to 2003 time era. There is in particular certain Statistical Sightings of Better Angels to report on. In particular, using change-point analysis methods worked out in a Cunen, Hermansen, Hjort 2017 paper, a confidence curve for the potential Better Angels change-point, where the degree of interstate war violence can be seen to go a bit down, compared to before, indicates that 1965 is such a breakpoint. Interestingly, the world famous Steven Pinker instantly liked the blog post, and tweeted, to his 367001 followers: "Sophisticated new analysis of the stats of war by Nils Lid Hjort affirms a decline over time, but finds the sharpest in 1965, nor 1945." Admittedly, he doesn't go quite as far as "The most inspiring blog post I've ever read", but he does say: "It’s a fascinating and sophisticated analysis, beautifully presented. It is interesting to consider 1965 as a possible breakpoint for a historical decline war. Certainly one finds mainstream attitudes changing around then, with an unprecedented (I think) antiwar movement; I suspect an analysis of cultural motifs (perhaps with Google ngrams) would show a transition starting around then."
So, in view of the statistical modelling carried out in the blog post, with the tentative conclusions drawn – wouldn't it be nice if, as the B.B. so eloquently express it, indeed from precisely that point on our common species time axis, if the Vietnam War, identified in the Cunen-Hjort change-point plot above, quickly followed by the Flower Power of Sergeant Pepper and Woodstock, really turns out to be the Beneficial Breakpoint, the Charitable Change-point, the Generous Game-changer, the Radical Regime-shift, complete with a Before and an After attached to it, where the fundamentally cute but occasionally terrifying homo sapiens species has managed to become, well, significantly less warrying and violent and belligerent, than before?
Statistics of Tolstoyan Ambitions: War-and-Peace Analysis, PRIO January 2018
The Peace Research Institute of Oslo (PRIO) organises a select series of Oslo Lectures on Peace and Conflict, and on January 11 the invited speaker was Aaron Clauset. The ensuing panel discussion, with Håvard Hegre, Nils Lid Hjort and Siri Aas Rustad, paid particular attention to the partly controversial question mark in Aaron's talk title Towards a More Peaceful World?. Here Nils, using a Cunen-Hermansen-Hjort method for constructing confidence curves for change-points and regime shifts, is discovering the Vietnam War – the answer to one of several versions of a basic question, if there is a change-point in the history of wars, from 1823 to 2005, in the statistical distribution of battle deaths, with higher mean to the left and a smaller mean to the right.
Part of the background here is that Nils will supervise Jens Kristoffer Haug, a new master student, jointly with Håvard Nygård at PRIO, with a project on such Tolstoyan proportions, modelling, analysing and interpeting war-and-peace data. This has also occasioned an interview and story in titan.uio.no, relating also to official higher-level hopes for collaboration between the University of Oslo and PRIO. Nils is reading Война и мир, with Nygård and Haug reading The Long Chapter Five (on The Long Peace) in Pinsker's classic The Better Angels of Ouc Nature.
Whales, Politics, and Statisticians: another FocuStat Blog post, January 2018
To produce a mighty book [or perhaps a mighty blog post], you must choose a mighty theme. No great and enduring volume can ever be written on the flea, though many there be who have tried it. Says Melville, in Moby-Dick; or, The Whale. And in our statistical working lives, clashing with whales and international politics decidely counts among our mightier experiences.
During the spring of 2017, Céline Cunen (in front) and Nils Lid Hjort (rowing) were offered a peek into the exclusive and partly politicised world of whaling research. We were invited to report to and attend the May 2017 meetings of the Scientific Committee of the International Whaling Commission. This turned out to involve whales, biology, developing FIC machinery and confidence curves for whales, history, power politics, and more. In our longer-than-usual blog post account we describe various features of these experiences, and also have a stab at a somewhat common phenomenon when statistics clashes with politics: One group claims that "a fact has been proven", whereas an opposing group says "no, there is so much surrounding uncertainty that we can't be sure yet" (etc.).
FocuStats Julebord og Generalforsamling, December 2017
We've been more than reasonably though not yet unreasonably socially constructive & creative, since we first established something resembling the FocuStat group to come, in June 2014. We've had workshops and research kitchens and travelled to conferences and had guests – and we've had julebord, each December, and spring semester parties, each June. On previous occasions we've been at the places of Céline, of Gudmund, og Kristoffer, of Nils, and now we went to Emil's place. In spite of or because of a certain clair-obscur style we touched on a wide variety of topics, as usual; we also sang a four-part motet or two, distributed Christmas presents, and gave speeches. The clair-obscured Kristoffer, Céline, Emil, Gudmund, Nils, Sam-Erik also toasted the absentees Riccardo and Vinnie. And yes, we made various bold plans for A.D. 2018.
Centre of Excellent Jule-, Juletre- og Julenisseforskning
Nils Lid Hjort is apparently a member of the Senter for Fremragende Jule- og Julenisseforskning at the Faculty of Mathematics and Natural Sciences – "Senter for Fremragende Jule- og Julenisseforskning ble etablert i 2015 av Gaute Einevoll, Sunniva Rose og Anders Hafreager, som senere har fått følge av Nils Lid Hjort og Klaus Høiland", reports Titan. These researchers have apparently specialised in innovative, cross-cultural and cross-disciplinary research on Christmas, jolesveinar, Santa, grøderikdomsgudar, et al. We've earlier, actually on more than one occasion, disproven earlier so-called proofs of Santa's non-existence ("to clear up this important question, Internet magazine forskning.no has gathered four of the country’s sharpest researchers: astrophysicist Knut Jørgen Røed Ødegaard, professor of physics Gaute Einevoll, professor of mathematics Nils Lid Hjort and Elf expert Ane Ohrvik"). On the contrary, we've argued the contrary!, i.e. his existence. Nils has also been expressing his pro-Christmas views via singing this December -- and he still stands by his Deilig er Jorden essay in the Ad Fontes book (deilig er jorden, og det er det som er den egentlige provokasjon).
Classics: Nils Lid's "Jolesveinar og grøderikdomsgudar" (1933) and "Um dualisme og festtider" (1935), side by side with one of the more irrefutable proofs of Santa's existence, the Carl Barks 1949 story Letter to Santa.
Three more FocuStat blog posts, December 2017
The FocuStat blog is well and alive, with Statistical Stories of different kinds for the Concerned and Curious Society, with three more posts in December 2017 – and with yet more stories to come. Céline Cunen, Gudmund Hermansen, Emil Stoltenberg offer a detailed book report for the Schweder and Hjort book, Confidence, Likelihood, Probability (Cambridge University Press, 2016); Tore Schweder writes about certain frequentist misgivings to the Bayesian type analyses propagandised by C. Sims (recepient of the Sveriges riksbanks pris i ekonomisk vetenskap til Alfred Nobels minne), and Nils Lid Hjort explains why he chooses to disagree, in part, with about 35 authors, regarding the analysis and interpretation of data from a JAMA October 2017 study of whether the therepeutic thermia intervention (the so-called cooling of newborns, in extreme cases of oxygen deprivation during birth) is still beneficial to neonates, if initiated more than six hours after birth. Nils also forced the 6 PhD and 19 Master level students in his STK 4021-9021 Bayesian Analysis course through a somewhat strenuous ten-day exam project with a list of questions pertaining to precisely this JAMA October 2017 article.
Two more Master students finish their work, December 2017
On December 11 2017, two FocuStat master students, both supervised by Nils Lid Hjort and both having research themes based on the CLP course, had their final examinations, successfully defending their theses! Josephina Argyrou wrote on "Topics in Confidence Distributions", with emphasis on t- and t-boot-constructions, whereas Leiv Tore Salte Rønneberg's thesis was "Fiducial and Objective Bayesian Inference".
From Processes to Models: Research Kitchen, November 2017
We've had a small series of very fruitful and indeed kitchenish Research Kitchens: Empirical Likelihood (September '14), Minimum Divergence and Scoring Rules (September '15), L^\eta, Многая лета (October '16), and now From Processes to Models (November '17). We kitchened away, in Hurricane, Danmarksbåten, Gates of Eden, and in VB Room 123. On the third day of the kitchen we took time to listen to Haydn (1732-1809) and Mendelssohn (1809-1847), illustrating some of the many statistical concepts and ideas floating around; how to make two-stage cumulative influence processes into flowing operations, with the second component positively dependent upon the first, and with a cure at the end. Participants giving presentations were Bo Lindqvist, Solveig Engebretsen, Thordis Thorarinsdottir, Martin Tveten, Nils Lid Hjort, Riccardo de Bin, Céline Cunen, Claudio Heinrich, Jonas Moss, Emil Stoltenberg, Riccardo Parviero. Themes included repair of complex systems, pain tolerance, the climate over the next century, modelling our planet, the Wars of Roses, how to avoid Titanic sinking, the lives of Roman Era Egyptians, aggressiveness level following violent videos, how medical histories of pregnant women influence their children, and, of course, basketball.
Céline Cunen's FocuStat Mediaeval History Blog Post on the Wars of Roses has been picked up by media in at least twelve languages
Céline Cunen (left) is becoming world famous, apparently. Her FocuStat Blog Post Mortality and Nobility in the Wars of the Roses and Game of Thrones (posted on our website October 2015) has attracted world-wide attention! She has given several talks and interviews about the relevant statistical, historical, science-fictional themes, and different types of media, in at least a dozen different languages, have discussed her work. Not bad for a PhD student. Jeg er en del av en forskergruppe som ønsker å formidle statistikk. Dette var en mulighet til det! Cunen is a statistician, and a large part of her job is to find sample selections that give a meaningful basis of comparison. Cunen sieht aber auch klare Unterschiede zwischen Serie und Realität; sie hebt in ihren Studien insbesondere die Staffeln 1-5 als besonders akkurat hervor. En la investigación, Cunen descubrió que la historia de Martin refleja de manera muy similar a la realidad el nivel de muertes de personajes nobles que fallecieron en la Guerra de las Rosas, incluso se ve marcado en la cantidad de mujeres y hombres que perdieron sus vidas.“La guerra delle due rose è la principale fonte di ispirazione per GoT e mentre lavoravo a una delle mie lezioni di dottorato, mi è venuta l’idea di esaminare se la guerra della serie tv fosse stata più sanguinosa di quella reale”, spiega Cunen. Ph.d.-studerende Cunen fra Universitetet i Oslo har undersøgt, om dødsraten i serien nu er så forskellig endda fra de engelske borgerkrige mellem 1455 og 1487 kaldet Rosekrigerne, som forfatteren George R. R. Martin selv har udtalt, at hans bøger og seriens firlæg er baseret på. Det er den ikke. Cunen, doktorantka na wydziale matematyki na Uniwersytecie w Oslo, porównała wskaźnik śmiertelności postaci w serialu HBO ze średnią długością życia angielskiej ludności podczas Wojny Dwóch Róż. Cunen, doktorale student in wiskunde aan die Universiteit van Olso, het GoT vergelyk met die War of the Roses-oorlog. 那工夫哪有這種東西」之類的職業闢好。假設能把影劇中所看到的視覺影像與本身對付數字之專業相共同，跨界運用到戲劇和汗青研討舉行對話，那又是別的一種差別的夢境條理。挪威奧斯陸大學（Oslo University ). Аспирантка Университета Осло Селин Канен сопоставила показатели смертности в «Игре престолов» и войне Алой и Белой розы — серии важнейших династических конфликтов Англии XV века.
Inference With Confidence: Special Issue of the JSPI, with Hjort and Schweder as Guest Editors (November 2017)
In May 2015, FocuStat organised the international workshop Inference WIth Confidence, with focus on theory and applications for confidence distributions and confidence curves, and with invited talks from various scholars. This led also to an agreement with the Journal of Statistical Planning and Inference that a Special Issue should be published, with Nils Lid Hjort and Tore Schweder as Guest Editors, on Confidence Distributions and Related Themes. After various efforts, by patient authors, referees, and colleagues, along with the customary revision processes, this has resulted in a first-rate collection of eleven journal articles, in addition to a general introduction article by the guest editors, published in November 2017. Some of these articles have been based on presentations first given at the FocuStat Workshop in May 2015, but other top scholars in the field, not present in the Oslo Workshop, were also recruited.
Sylvia Richarson given honorary doctoral degree at the University of Oslo, September 2017
Professor Sylvia Richardson, Professor of Biostatistics, Cambridge University, and also visionary and inspiring Director of the Medical Research Council Biostatistics Unit, Cambridge, is a good colleague of several of the UiO statisticians. She was one of 16 new honorary doctoral degree holders in connection with the UiO celebrations 1st of September, 2017 (such honoris causa celebrations take place only every third year). The UiO statisticians, with full backing from the Departments of Mathematics and Biostatistics, used the occasion to organise a full Sylvia Richardson Day, where several FocuStat people also contributed. Nils Lid Hjort gave a talk with the unusually long title "All models are wrong, but some are more biologically plausible than others: Survival analysis models based on crossing times for cumulative damage processes". In a cabaret of seven young & up-and-coming statisticians, Céline Cunen, Emil Stoltenberg and Kristoffer Hellton contributed presentations, on respectively "Game of Thrones vs Wars of Roses: Survival and competing risks", "Some insights into the cure model", and "Personalized prediction through focused fine-tuning of ridge regression". The photo, taken by the cellist of the Piatti Quartet, is from the Grand Festivities at Grand Hotel on September 1, showing our new Honoris Causa along with proud colleagues Sven Ove Samuelsen, Ingrid Glad, Arnoldo Frigessi, Nils Lid Hjort. Examination appears to reveal that approximately two thirds of the male professors managed to accurately follow the dress-code for the occasion.
Céline Cunen gives invited talk at the 61th ISI World Congress in Marrakech, July 2017
Her invited talk, in loco parentis, was "Combining diverse information sources with the II-CC-FF paradigm: Independent Inspection, Confidence Conversion, Focused Fusion" (joint work with Nils Lid Hjort), in the invited session on "Bridging BFF (Bayesian/frequentist/fiducial) inferences in the era of data science". The other speakers in the session were Jan Hannig and Robin Gong, with Min-ge Xie as discussant (these are all in the international and steadily widening FocuStat Friends club). This 61st ISI World Conference took place under the high patronage of His Majesty King Mohammed VI!
The 11th Bayesian Nonparametrics Conference, Paris, June 2017
Céline Cunen, Gudmund Hermansen and Nils Lid Hjort took part in the 11th Bayesian Nonparametric Conference, held at École normale supérieure, Paris, June 25-30, 2017, along with some 150 other participants from a wide range of nations. As Nils reminded the audience, on page 1 in his invited talk, the previous such BNP conferences have taken place in Belgirate '97, Reading '99, Ann Arbor '01, Roma '04. Jeju '06, Cambridge '07, Torino '09, Veracruz '11, Amsterdam '13, Raleigh '15 (the only person who has managed to take part in all of these is Peter Müller, whereas Nils has been present, and giving invited talks, in n - 1 of these).
Nils spoke on the empirical likelihood coupled with Bayes. Gudmund gave an inspired poster on Bayesian nonparametrics for stationary time series, and was a "hit", having to explain and defend himself and his methods for three hours of wine and shouting, with steadily new eager individuals and groups coming to see what was going on. His poster made him a FavourableMentionee from the jury (senior BNPers Wes Johnson, Catherine Forbes, Steve MacEachern), who scrutinised each of the 106 posters. [Photo credit: Christian Robert, one of the chief professors, both in the field and in Paris statistics generally -- and since BNP11 was a low-cost affair, Christian and the other top professors had to do all the dishes. In Robert's heavily visited Xian Blog, on statistics and Bayes and climing and more, he was kind enough to characterise Nils's and Art Owen's talks as "absolutely fantastic".] Some more photos and comments are on our Facebook pages.
Det 19. norske statistikermøtet i Fredrikstad, June 2017
Here we are if not at the top of the world, then at least at the top of Fredrikstad -- the full group (from left: Walker, Hermansen, Ko, Stoltenberg, Cunen, Hjort, Hellton; photo: Jonas Christoffer Lindstrøm) took eagerly and near dominantly part in the 19th Norwegian meeting of the Norsk statistisk forening, in Fredrikstad, June 12-15, 2017.
For the talks we gave, check under Talks on this website (somewhat astoundingly, we've given 28 talks, so far, from January 2017 to June 2017). There were also talks given by various FocuStat friends and associates: Kathrine Frey Frøslie and Per Mykland gave specially invited talks, and former Nils students Steffen Grønneberg, Jonas Moss, Martin Jullum gave presentations. The Sverdrup Prize 2017 was awarded Ørnulf Borgan, where the Adelskalender so far lists Tjøstheim (2009), Schweder (2011), Hjort (2013), Aalen (2015), and now Borgan (2017). For some photos and further comments & views, check our Facebook Album for Fredrikstad.
FocuStat Workshop May 2017: Building Bridges (at Bislett)
Like for our previous workshops (Inference With Confidence, May 2015 and FICology, May 2016), our BBB Workshop (May 22-23-24, 2017) took place at Teknologihuset, a javelin's throw away from Bislett Stadion. The themes, broadly speaking, involve ways of inventing, assessing, examining and fruitfully utilising methods for combining parametrics with semi- and nonparametrics. In addition to various talks by the FocuStaters and several Oslo colleagues, we had the Bergen trio of Tjøstheim, Otneim, Støve with Hjort-Jones approaches to higher-dimensional density estimation and Gaussian correlation functions; Sonia Petrone and Igor Prünster from Milano with Bayesian nonparametrics in their bags; Thore Egeland with genetics, Christian Rohrbeck with spatial modelling, etc. Our excursion was to Observatoriet, the oldest University of Oslo building, where the participants were properly morleyed. For some photos and further commentary, see our Facebook pages.
Céline Cunen gives presentations to the Scientific Committee of the International Whaling Commission, May 2017
Celine Cunen and Nils Lid Hjort took part in the 67th Meeting of the Scientific Committee of the International Whaling Commisson, in Bled, Slovenia, May 10-14, 2017. On May 11, Celine gave as many as three presentations to the Scientific Committee, based on work with Nils and senior scientist Lars Walløe (who has been the Head of the Norwegian Delegation to the SC of the IWC for many years, scientific adviser to the Norwegian government, etc. etc.). (More text to come here once the meeting is finished.)
The photo shows Johan Hjort in field work in 1902. He was among the driving forces behind the creating of the IWC.
Bayes, Fiducial, Frequentist; Best Friends Forever: BFF4 at Harvard, May 2017
The BFF series of conferences is growing in size, depth and importance. In April 2016, some of us took part in the BFF3 at Rutgers, and in May 2017 Nils participated in BFF4 at Harvard, giving an invited talk on data fusion based on confidence curves, based on joint work with Celine. Taking part in this conference was enjoyable and also a rare opportunity to see a crucially important old-ish segment of The History Book of Statistics, a segment which will not be there in a dozen years. We saw Sir David Cox (b. 1924) give a splendid talk by skype; Don Fraser (b. 1925) is as always inspiredly present; I thanked Herman Chernoff (b. 1923) again for the icecream he bought me in 1991, and he told me his first paper was published in 1946; Art Dempster (b. 1929) told us his belief functions; we listened eagerly to (and conversed with) a range of other clever-and-sharp pioneers (Terry Seidenfeld, Don Rubin, Sandy Zabell [on I.J. Good's secret life], Larry Brown, Nanny Wermuth, etc.). In the later segment of Jim Berger, Nancy Reid, Vijay Nair, Xiao-Li Meng, Christian Robert et al., we felt, well, young, in comparison, but still in awe of the intellectual achievements and egg-sharp presence of those upon whose still visible shoulders we stand. Also remarkable, for this BFF4 conference, was the active participation of several philosophers.
Soumission & Submission (March 2017)
[From left: Gudmund, Emil, Vinnie, Sam-Erik, B. Cuda, Céline, Kristoffer, Nils.]
Betwixt submission of journal articles, we took soumission time for a different type of active, social evening. At our end-of-semester parties, Nils famously forces his compatriots to break out in song -- well, in organised, four-part arrangement terms, complete with instructions regarding prior individual rehearsals. On this occasion, 31st of March, we devoted an evening to discuss Houellebecq's 2015 novel Soumission. We had all read the book before the meeting, of course (Céline and Emil in French, Vinnie in Dutch, the rest of us Norwegian and English editions). Some of us are now looking forward to the Bayesian Nonparametrics conference in June, near Sorbonne, in the middle of Soumission District, with a bit of apprehension.
Céline Cunen gives the statistics miniforelesning for Åpen dag, March 2017
About once a year the UiO Blindern campus is crawling with hopeful & prospective future students, from videregående skole and elsewhere, to the Åpen dag event. So on March 9 these youngsters had a chance to sample from an impressive range of fine lectures from various prominent (!) lecturers. Céline Cunen was elected to give the statistics talk, and her topic was "på død og liv i Game of Thrones", partly modelled on her FocuStat blog post on these themes. Her GoT work, comparing GoT life survival with that of the Wars of the Roses (1455 - 1487), has earlier been featured in both Universitas and Titan, and then, intriguingly, picked up by various groups in various fora in other countries -- ein derartiges Massenableben wie in Game of Thrones sei aber durchaus realistisch für die mittelalterliche Zeit der Geschehnisse. Das stellte Celine Cunen, Akademikerin an der University of Oslo, im Rahmen einer Studie fest; Sa conclusion est que Game of Thrones a un ratio plutôt normal. Elle a réalisé que le nombre de morts parmi les nobles de la série est une juste représentation de ce qui s’est passé durant la Guerre des Roses; Para simplificar las cosas, Cunen realizó un estudio estadístico en el que creó una base de datos con las personas que se involucraron en la guerra civil inglesa; 《冰與火之歌：權益遊戲》劇中的死得透澈率可否真實？」（The death rate in Game of Thrones is realistic）？這一題旨乃是奧斯陸大學數學系博士生 Celine Marie Løken Cunen 現在正在舉行的一項計畫，實行用數學調查來闡發《冰與火之歌：權益遊戲》，劇中呈現的視覺刻畫可否可靠; along with reader comments etc.
Det Kongelige Norske Videnskabers Selskab & a few more talks
Nils Lid Hjort is now an elected member of Det Kongelige Norske Videnskabers Selskap (the Royal Norwegian Society of Sciences and Letters), since January 2017. This Trondheim based learned society, founded in 1760, is a few generations older than the perhaps slightly more well-known Det Norske Videnskaps-Akademi (the Norwegian Academy of Science and Letters), founded in 1857, and where Nils became an elected member in 1999. -- Kristoffer Hellton gave a pop. math. talk on the seven pillars of statistical wisdom, to an eager crowd of mostly UiO MatNat-Fakultet students, on January 26. Nils gave a gentle introduction to Bayesian Nonparametrics as part of the series of Big Insight Lunch Talks, in February 1.
Geilo Winter School, January 2017: Machine Learning
The Geilo Winter Schools have been an annual affair since 2001. This year's programme was on Machine Learning, and attracted a record number of 158 participants, most of whom PhDs and PostDocs from Norway and neighbouring countries, but also interested parties from more distant countries. From our group Céline Cunen, Gudmund Hermansen, Nils Lid Hjort and Emil Aas Stoltenburg took part. Nils gave a total of six hours of lectures, with the main components being Model selection and model averaging (2 hrs), Confidence distributions and data fusion (2 hrs), and Bayesian nonparametrics (2 hrs). Céline gave a poster presentation. Some of us also skied all the way up to Prestholt (1242 m.o.h.), close to wondrous Hallingskarvet. Overall the five-day school testified to the broad and active interest in Machine Learning and associated statistical applications of Data Science in Norway anno 2017, from both theoretical, applied and industial perspectives.
Statistical methods for mediaeval mysteries: Céline in Titan & Tirant lo Blog
Céline Cunen is apparently one of Norway's more mediaeval statisticians. In addition to work on the Wars of Roses (1455-1487) she is now featured in Titan with an article on the adventures of Tirant lo Blanch. This is arguably the world's first ever novel, written in Catalan, around 1460, about 150 years before Don Quijote. The mystery in question, which we attack using methods of confidence curves, etc., is to understand where in the book's 487 chapters Author Two (Martí Joan de Galba) took over for Author One (Joanot Martorell). There's also an accompanying FocuStat blog post, fleshing out more of the methodological details. The background paper is Confidence distributions for change-points and regime shifts, by Céline, Gudmund H and Nils LH, to appear in Journal of Statistical Planning and Inference. Other applications include the Hjort liver quality time series (1859-2013) and the number of skiing days per winter at Bjørnholt. (Nils LH is incidentally a direct descendant of the Titans -- Ōkeanós Titan, who mingled with the humans some 3500 years ago, is his 88th great grandfather.)
Met hartelijke groente, Emil Stoltenberg in Dax18, fagligpedagogisk Céline Cunen, November 2016
The FocuStat-ers are occasionally visible even outside our own corridors. Vinnie Ko, Neederlandse schrijver, columnist, wiskundige, statisticus, and Nils's most recent PhD student, is busy this month in connection with his book Met hartelijke groente: Nederlands door Koreanse oogen being released. 48 hours before Trump-Clinton, Emil Stoltenbeg helped the Norwegians understand various complex aspects of the art of predicting the American presidential election, in the Dax18 tv programme, leaving us all wiser. Céline Cunen gave a "popular science" lecture at the faglig-pedagogisk day, comparing survival distributions for the Wars of the Roses (1455-1487) with the Game of Thrones universe. A few weeks back, Nils Lid Hjort did his third doctoral degree attack this year, this time attacking Håkon Otneim at his disputation in Bergen.
L^\eta, Многая лета, research kitchen October 2016
Our research kitchens are smaller-scale meetings, where the emphasis is on presenting, reacting, discussing (including elements of constructive disagreements!), as opposed to the usual conference or workshop format. In 2014 we did empirical likelihood and cure models, in 2015 divergence based estimators and scoring rules, and on this occasion we did L^\eta, Многая лета. Themes involved constructions where the likelihood function is not used directly in full force, but raised to a factor typically smaller than 1, for reasons of Bayesian or frequentist robustness, regularisation or penalisation, etc. We judged it a success, with a vigorous atmosphere of learning, exchanging views, causing some of us to jump to the blackboard, chalking away, and getting inspired for more work (!). The presenters were Peter Grünwald, Peter Müller, Nils Lid Hjort, Tamara Broderick, Per Mykland, Gudmund Hermansen, Ingrid Glad, Jeff Miller, Sam-Erik Walker, Kristoffer Hellton.
The Klækken PhD and PostDoc seminar September 2016
The annual PhD and PostDoc seminar at Klækken Hotell took place September 9-10. Céline Cunen, Kristoffer Hellton and Emil Stoltenberg were among those giving presentations based on their research work. Céline's theme was the world's first novel (Tirant lo Blanc, c. 1460), and she is progressively mediaevel, or at least seen as such by the press, now including Natt og Dag, which the same week featured a brief piece on her work related to Game of Thrones and the Wars of Roses: "Cunen har påvist at det er flere likheter mellom Game of Thrones og middelalderen. Bra påvist! Påvis mer!" Kristoffer talked about Fridge and Emil about alternatives to the Cox regression model for certain problems with censored data.
Game of Thrones, the Wars of the Roses, and the World Championships of Pre-Orienteering
Some of us are sufficiently interesting to occasionally land in the news. Céline Cunen has been interviewed (again!) in the University of Oslo student newspaper Universitas, about her statistical analyses of life lengths and death rates in the two universes of Games of Thrones and the Wars of the Roses; see also her blogg post on these issues. Apparently, the underlying hazard rates, for these two populations, share certain similarities, though they refuse to follow Cox's proportional hazards model (!). Incidentally, our colleague and friend Sir David Spiegelhalter, "Professor Risk" at Cambridge, has done some prediction analyses and warned Britain that if they continue to watch the Game of Thrones, there'll be no sex in England, by the year 2030.
Martin Jullum has once again got himself some metal, this time the silver medal (just behind Martin Fredholm, Sweden), in this year's World Championships of Pre-Orienteering, held in Strömstad, Sweden, August 2016.
Post-Selection Inference Workshop in Leuven
Céline Cunen, Gudmund Hermansen and Nils Lid Hjort took part in an inspired & inspiring two-day workshop on post-selection inference, in Leuven, August 22-23. Nils gave an invited talk, on post-selection distributions, model averaging, and bagging. We spent a very pleasant evening at the home of Daumantas Bloznelis, who was our FocuStat guest on a certain occasion last year (look under Guests), who has since finished his PhD at NMBU Ås, and is now having a one-year PostDoc experience in Leuven.
Joint Statistical Meetings 2016 Chicago
The annual Joint Statistical Meetings are the biggest meetings of statisticians, typically with more than six thousand participants from over fifty countries, a huge number of invited and contributed sessions, etc. JSM2016 took place in Chicago, July 30 to August 4. Nils gave an invited talk on the II-CC-FF paradigm for effective combination of information across diverse sources (Independent Inspection, Confidence Conversion, Focused Fusion), and his applications included the statistical message that the International Olympic Committee really ought to let Chicago's own world record holder and Olympic gold winner in the 1000 m sprint (photo: Chicago Tribune) skate the 1000 m twice, in the future Olympics. There were several sessions on the BFF theme (Bayes, Fiducial, Bayes; Best Friends Forever), of relevance for part of the FocuStat work programme, and with many of "our friends" giving talks. There will also be BFF themes in the JSM2017, in Baltimore.
Céline Cunen and Kristoffer Hellton win Best Poster Awards at NordStat2016 København
The FocuStat core members intend to have at least one conference trip each year where we all participate. In July 2015, we went to Rio. For 2016 the group decision was to go to NordStat2016, the Nordic Meeting on Statistics, in København (and next year we may up the ante further, going to either Fredrikstad or Marrakesh). Céline Cunen, Sam-Erik Walker, Gudmund Hermansen, Kristoffer Hellton gave poster presentations and flash talks, while Nils Lid Hjort organised a session on Statistical Theory (and gave a talk on hybrid likelihood). Among around sixty different high-quality posters, the three winners of the Best Poster Awards were Céline, Kristoffer and home favourite Frederik Riis Mikkelsen.
Nils has been opponent at the PhD disputation defense of Lars-Erik Gangsei, May 24, at the NMBU, Ås; the PhD thesis in question was "Linear Multiresponse Models -- Theoretical Developments and Applications in Porcine". During the attack & defense no particular reference was made to Nils Lid's doctoral degree "Norske slakterskikkar, med jamføringar frå nærskylde umråde". Nils was also the opponent for the licentiate degree of Ivar Simonsson, Chalmers University, Göteborg, June 3; the thesis was "Bayesian networks: exact inference and applications in forensic statistics".
FocuStat Workshop May 2016: FICology
Our first international FocuStat Workshop was Inference With Confidence, in May 2015, on confidence distributions and related themes. Our second international workshop will be held Mon-Tue-Wed 9-10-11 2016, on FICology (click here or on the summer Bislett image to the right for the programme). These Survival of the FICest themes relate to using focused information criteria for model building, model selection and model averaging, in theory and applications (cf. the Claeskens and Hjort book Model Selection and Model Averaging, Cambridge, 2008). Invited speakers include Gerda Claeskens, Holger Dette, Axel Gandy, Peter Grünwald, Ingrid Hobæk Haff, Håvard Hegre and/or Håvard Nygård, Thijs van Ommen, Ingrid Van Keilegom, Lars Walløe, along with the FocuStat locals and some further FocuStat-associated statisticians, like Steffen Grønneberg, Martin Jullum and Peder Østbye. If you take an interest in these themes, regarding FICology theory or applications, and would like to contribute or attend, please contact Nils LH or Céline C. The workshop will take place at Teknologihuset, about a javelin's throw from Bislett (though the number of workshop participants is likely to be less than 28,000).
Short course on confidence distributions
Nils LH gave a ten-hour short course at the Bicocca Universy of Milano, April 19-22, on confidence distributions and related themes, based on the Confidence, Likelihood, Probability book, to an appropriately fascinated forum of PhD students and faculty. He also visited Bocconi University and gave a talk there on confidence curves for change points and regime shifts (based on the Cunen, Hermansen, Hjort paper).
BFF: Bayes, Frequentist, Fiducial (Best Friends Forever)
The third BFF Conference was organised in April 2016 by the Department of Statistics and Biostatistics at Rutgers University, April 2016. The themes were foundational (what is a probability? are there different kinds?), far-ranging, inspirational, and with quite a few heavyweighters taking part (including Brad Efron, Don Fraser, Nancy Reid, Xiao-Li Meng, Jim Berger, Andrew Gelman, Jan Hannig, Regina Liu, Min-ge Xie, Ryan Martin, Ed George, Harry Crane, and Sir David Cox gave a 45 minute lecture sitting in his armchair in Oxford). Céline gave a poster and Nils an invited lecture, both relating to confidence distributions for change points and regime shifts.
Schweder and Hjort: Confidence, Likelihood, Probability
Tore and Nils's CLP 500-page book is (finally) out, Cambridge University Press, March 2016, as part of their flagship series Cambridge Series in Statistical and Probabilistic Mathematics. For some words of description and some praise (?), check "this lively book". Nils LH gives a course on this material at the Department of Mathematics, STK 4180: Confidence Distributions.
Martin Jullum's PhD disputation, April 2016
Martin Jullum's dissertation "New focused approaches to topics within model selection and approximate Bayesian inversion", with Nils LH and Odd Kolbjørnsen as supervisors, has been found satisfactory by the PhD committee, consisting of Ingrid Van Keilegom (UC de Louvain), Jo Eidsvik (NTNU), and Anders Rygh Swensen. The public defence & disputation took place Friday April 1st, and Martin's "given topic" for the occasion was Empirical Likelihood.
Nils talks to kids (and to some grownups)
The so-called Humanistforbundet at the University of Oslo organises a string of mini-lectures where UiO professors talk to a select group of fourteen-fifteen years old school pupils, about their fields of interest and how these might relate to the society at large, and, if possible, how their profession relates to themes of solidarity. The point is also to "get the kids engaged". The ungdomsundervisningen hence aims at providing bits & pieces of influence and exposure to the world of academics, for these youngsters, in a spirit perhaps close to the "konfirmasjon" process. Nils LH contributed his two-hour part in this process, in February 2016, talking about "statistikk, sannsynlighetsteori, sjansespill, samfunn, solidaritet", and yes, the kids responded, more than adequately. -- For a slightly different segment of people and related to a more than slightly different type of work, Nils LH is now on the steering board of the Norwegian Academy of Science and Letters, being "nestleder" of the Natural Sciences Division of the Academy.
Three new master students
As many as three new master-level students are starting out this spring semester 2016, with Nils LH as supervisor; details will be given later, and their names will honorably appear in the Who We Are list. Their research themes will probably relate to confidence distributions and model selection.
Science, ethics, society, and exams
PhD students at the Faculty of Mathematics and Natural Sciences are required to take 5-credit course on science, ethics and society (intriguingly, these official English words do not quite accurately reflect the official Norwegian words for the course, "forskningsetikk, vitenskapsteori og vitenskapsmetode"). The exam consists in writing up an essay on a relevant topic. Céline C's essay is The impact of Philosophy of Science: A citation analysis, whereas Sam-Erik W's is Noen etiske problemstillinger i forbindelse med håndtering og formidling av usikkerhet i miljøforskning (both from December 2015). Both passed with flying colours.
Another Master & the Sirens of Titan
Jonas Moss has completed his master thesis Cube Root Asymptotics in November 2015, supervised by Nils LH, and officially earned his Master's degree on December 15. His supervisor was very satisfied, and so were the markers. -- In December 2015, the Faculty of Mathematics and Natural Sciences heralded & launched its ambitious web-magazine Titan, and in its first week there were two stories concerning statistics, both featuring "our" Sirens of Titan. Céline C, who a few weeks earlier spoke on Tirant lo Blanch of the 1460ies is becoming progressively mediaeval, and is interviewed about her work comparing survival distributions of Game of Thrones with the Wars of the Roses; see her FocuStat blog post. Our other siren is our group's close friend and collaborator Ingrid Glad, who was Titan's professor of the month.
Velferdsprisen (the Cultural and Social Activities Award) 2015
"The Cultural and Social Activities Award is given to a group of members of staff, an academic group or an individual who has made an outstanding contribution to enhancing employee satisfaction and a sense of community at the University of Oslo – for example, by taking a new approach to conditions in the work place, improving the physical environment or spreading enthusiasm and satisfaction in other ways." No, admittedly, the FocuStat group hasn't won the Velferdsprisen for 2015, but the university choir Unikor did, perhaps in some modest part due to the nomination & documentation prose, which Nils qua board member helped compose. Perhaps FocuStat should strive to win in 2016, with Unikor singing.
Pub med professor (and a PhD student): Analysing Tirant lo Blanch 1490
In academics there are many ways to report your findings, but this was a new variant for us. On November 26, Nils and Céline gave a stereo talk in Realistforeningen's Pub With Professor (and a PhD student) series, about the world's first ever novel (filled with surprising amounts of sex & drugs & power politics, an almost counterfactual novel following the Fall of Konstantinopel 1453), written 1460-1463 and published in collected form in Valencia 1490. The first author Joanot Martorell died somewhere in the process of writing and publication, after which the second author Martí Joan de Galba took over -- but nobody knows precisely where and when. We examine certain statistical characteristics for the book's 487 chapters and attempt to pinpoint where the change took place, along with a full confidence distribution for assessing the uncertainty of our estimate. En un mundo en guerra el arma más poderosa es la virginidad de una princesa (indeed). A pdf for the stereo talk is here.
Faglitterært forfatterstipend to Fodnæs, Grønneberg, Hermansen
Njål Foldnes, Steffen Grønneberg and Gudmund Hermansen have won 100,000 kroner!, namely the publishing house Cappelen-Damm's special stipend for the most promising & fruitful book project in the field of higher education in Norway. They will write a book on statistics for bachelor students of economics, administration and marketing (with "En moderne introduksjon til statistikk for økonomi og markedsføring" as working title). Not all bank cheques measure more than 1 m^2, but this one does.
Alan Gelfand Week, October 2015
We've had a pretty intense week with Alan Gelfand, October 13-16. Alan is the James B. Duke Professor of Statistics and Decision Sciences at Duke University and among the international stars of the statistical circuit, with specialties including Bayesian statistics and space-time-modelling, and is here to be our "critical friend", looking over our shoulders to see what and how we're doing. He also gave a statistics seminar about Gaussian processes and their happy marriage to spatial data.
Audun Mathias Øygard completed his master thesis Measuring Similarity of Classified Advertisements Using Images and Text in September 2015, supervised by Gudmund HH and Nils LH, and officially earned his Master's degree on October 1. His markers liked it!, and so do finn.no and various other interested parties. Audun gave a talk about his methods and applications, with emphasis on "deep learning", at the Oslo Data Science Meetup 5th of October. This Meetup, with a sequence of meetings for the eager crowd of machine learners et al. in the Oslo region, also outside the "academic circles", is organised by Gudmund H.
Research Kitchen September 2015
Our first Research Kitchen was held in September 2014, with empirical likelihood, cure models and a bit of semiparametric modelling (see the 2014 section below). Our second Research Kitchen was held September 28-30 2015, with themes related to robust parametric inference, estimation and model selection via proper scoring rules and minimum divergence methods. Applications include statistical prediction and air pollution assessment. Our invited guests are Fabian Krüger from the Heidelberg Institute of Theoretical Studies, Monica Musio from Università degli Studi di Cagliari Matematica ed Informatica, and Thordis Thorarinsdottir from the Norwegian Computing Centre.
On the road: Emil in Prague, Martin in Biarritz
Emil Stoltenberg has been Norway's representative at the 19th European Young Statisticians Meeting in Prague, 31 Aug to 4 Sept 2015. He gave a talk based on his conference paper "The c-loss function: Balancing total and individual risks in the simultaneous estimation of Poisson means" with Nils, and their paper is available in the conference proceedings. Martin Jullum has taken part in the Petroleum Geostatistics conference in Biarritz, 7-11 Sept 2015, and gave a talk on "An approximation Bayesian inverstion framework based on local-Gaussian likelihoods", based on joint work with Odd Kolbjørnsen. Martin is also regularly "on the road" in more senses that this, and this July won bronze in the World Championships in pre-orienteering in Zagreb.
60th ISI World Congress in Rio de Janeiro, July 2015
All five of FocuStat's core group (Nils LH, Céline C, Sam-Erik W, Gudmund H, Kristoffer H) took active part in the ISI World Congress in Rio, July 2015. Nils gave an invited talk in a session on combining information across diverse sources, whereas Céline, Sam-Erik, Gudmund, Kristoffer gave presentations based on six-page contributed papers which will later appear in the proceedings volumes of the congress. See our list of talks and list of publications for details. Also, check out a little album of photos on our Facebook page.
Double Master Stoltenberg
Nils LH's student Emil Aas Stoltenberg has finished his second master's degree (June 2015), this time in statistics. His thesis title is The c-loss function: Balancing total and individual risk in the simultaneous estimation of Poisson means, and Nils and Emil plan to write up a paper this autumn, summarising the main findings. Emil's previous master's degree from 2013 is in political science.
Det 18. norske statistikarmøtet at Solstrand, June 2015
The FocuStat group had a strong presence at the 18th Norwegian Meeting of Statisticians, held at scenic Solstrand, June 2015, with four of us giving talks on the same day -- Céline C (Optimal inference for two-by-two tables modelled as Poisson pairs: a confidence distribution approach), Gudmund H (Predicting with confidence in time series processes), Nils LH (On just how surprised we ought to have been when Bolt set his world records), Martin (Parametric of nonparametric: The FIC approach for time series). Also, Tore Schweder gave an extended presentation on confidence disributions and epistemic probability, so June 17 was a day of FICology and CDs. Check the Facebook page for some photos and further comments. -- The photo shows the four Sverdrup Prize winners: Tore Schweder (2011), Dag Tjøstheim (2009), Nils Lid Hjort (2013), Odd Aalen (2015).
Klækken PhD and PostDoc Seminar
Each year, the statistics groups at the UiO Department of Mathematics and the Norwegian University of Life Sciences organise a two-day workshop for PhDs & PostDocs (and their supervisors), at Klækken hotell. At this year's Klækken seminar May 29-30, Céline C and Martin J gave talks based on their PhD work, respectively "Optimal inference for two-by-two tables: a confidence distribution approach" and "An approximate Bayesian geophysical inversion framework based on local-Gaussian likelihoods".
FocuStat Workshop May 2015: Inference With Confidence
An international-level workshop took place in Oslo, May 11-13, organised by FocuStat, on confidence distributions and related topics. Themes included construction of confidence distributions, performance issues, second-order corrections and fine-tuning methods, links to objective and empirical Bayes, using confidence distributions for combining information across diverse information sources, application stories, &cetera. The workshop was held at Teknologihuset, about a javelin's throw from Bislett. See details in the workshop programme. Queries might be directed to Nils LH or Céline C.
We're on Facebook
As of April 2015, Focustat has a Facebook forum page, meant for easy communication with friends & colleagues regarding news and events, along also with not-so-heavy-handed discussions of matters coming up, queries about various themes, tentative ideas worthy of being mentioned, etc. Look us up, check what's posted, and do not hesitate to contribute.
We've hired a second PostDoc
FocuStat has already on board one professor plus two PhDs plus one PostDoc (see the Who We Are section), and we have now got hold of a second PostDoc. The application deadline was in March, 2015, and from a strong field of competitors we are happy to welcome Kristoffer Hellton in our group. Kristoffer defended his PhD at the Institute of Basic Medical Sciences in February this year, with thesis title "On high-dimensional principal component analysis in genomics: consistency and robustness".
Norway win the Women's Handball European Championship 2014
On 21-Dec-2014, 1.7 million Norwegians and presumably a zillion other Europeans spellboundedly watched the grand finale Norway-Spain 28-25. But how exciting was the match, exactly? In precisely how dire straits did Norway find herself when Spain was leading 10-5 after 18 minutes, how relaxed should we have been on our sofas when Norway was leading 26-22 with 8 minutes to go, and should we have started trembling again three minutes later when Spain got back to 25-26? Nils has analysed the match and calculated precisely how excited we ought to have been, as a function of match time. These calculations can easily be made into an "app" providing the probabilities for winning and for a draw for the two teams playing, as the goals keep pouring in. A follow-up note will be published later.
Another PhD & another Master
On December 1, Jana Timková successfully defended her PhD dissertation Bayesian and Maximum Likelihood Nonparametric Estimation in the Monotone Aalen Model at the Faculty of Mathematics and Physics with the Univerzita Karlova in Prague. Nils has been co-supervisor for this project, and at least one joint paper will come out of these efforts. On December 17, Peder Østbye won his Master of Science degree here at the Faculty of Mathematics and Natural Sciences, University of Oslo, where Nils has been the supervisor. The thesis title is Cost Information-Value Trade-Off in Covariate Selection.
Ad Fontes, December 2014
A festschrift "Ad Fontes" has been published in connection with Terje Kvam's turning 70 in December 2014. The editors in question invited a group of "artists and writers" (and a bishop and a statistician) to contribute. Nils's essay for the occasion is Deilig er jorden.
And Quiet Does Not Flow the Don
Nils gives a talk in Trondheim 14/xi/14, concerning the potentially biggest plagiarism scandal in the history of literature, in the series of Pearls and Curiosa of Mathematics lectures.
Nils gives two lectures (and is on TV)
Nils gave a talk at the fagligpedagogisk dag 30/x/14 on Olympic cross-country skisprinting, proving, apparently, that the world is unfair for men, though not necessarily for women. He also gave a talk on 4/xi/14 on model comparison, model selection and model averaging at the SFI2 Farewell Workshop. He also had his three minutes of non-fame on NRK2 some time in late October, being interviewed by Jo Røislien in his "Siffer" programme, and tried to say something related to spontaneous, stress-related polydactylism and the octogonal number system.
The Johan Hjort Symposium
Of course Nils needed to take part in the Johan Hjort Symposium (Bergen, 7-9 October 2014), a conference on recruitment dynamics and stock variability held in connection with the 100 year anniversary of Johan Hjort's book Fluctuations in the Great Fisheries of Northern Europe, with about 150 participants from 23 countries. Nils gave a conference dinner speech about his great-grandfather. He hopes that he and Gudmund can write a paper analysing a certain Hjort time series 1859-2013 and its relation to Kola temperature trends over the last century, aiming for a special Hjort Issue of the Canadian Journal of Fisheries and Aquatic Sciences planned for 2015. -- Check also Nils's blogpost on this, with some personal reflections.
Our second PhD has started
Sam-Erik Walker is our project's second PhD stipendiat, starting from October 2014. He has a Master's degree in statistics from the Department of Mathematics, University of Oslo, from 2010, and has extensive experiences with research work from NILU, the Norwegian Institute for Air Research.
Gudmund Horn Hermansen wins his PhD ... and becomes our first PostDoc
On the 26th of September 2014, Gudmund successfully defended his thesis and obtained his Ph.D degree with the Department of Mathematics at University of Oslo. Congratulations! He didn't have much time resting on his laurels, as he immediately started working for FocuStat as our first PostDoc. In addition to his PhD work, on themes ranging from Bayesian nonparametrics to focused model selection and averaging for time series and nonstandard regression models, Gudmund has been research statistician with the Statistical Analysis of Natural Resource Data group at the Norwegian Computing Centre. He will maintain links with that research environment during his PostDoc period.
Research Kitchen September 2014
Our first Research Kitchen was held from the 2nd to the 8th of September. We were happy to welcome Professors Ingrid Van Keilegom (Catholic University of Louvain) and Ian McKeague (Columbia University). We discussed work related to Empirical likelihood and time-to-cure models.
Our first PhD has started
Céline Cunen is FocuStat's first PhD stipendiat, starting work at the department in August 2014. She has a Master's degree in statistics from Norwegian University of Life Sciences, related to partial least squares methodology for ranking and interaction detection with high-dimensional data.
The Norwegian Research Council funds ERC project for Nils Lid Hjort
This piece of "news" is from December 2013, but is recorded here for the appropriate historical reasons, the starting point of our FocuStat project. "Forskningsrådet finansierer Professor Nils Lid Hjorts prosjekt Focus Driven Statistical Inference With Complex Data. Matematisk institutt gratulerer!"