DSTS 50-års jubilæum - Wednesday

Back to Program (titles, speakers and times)

 

DSTS' historie

Niels Keiding (KU SUND)

Statistikfaget havde haft svært ved at etablere sig som disciplin i Danmark, men i foråret 1971 forelå den ene efter den anden præmis for at det var på tide at samle kræfterne:
Siden 1949 havde professor A. Hald (København) drevet Det Statistiske Seminar med (over 150) aftenforedrag og heldagsarrangementer af inden- og udenlandske eksperter. I 1958 oprettedes Eksamen i Statistik ved Københavns Universitet. Siden 1960 eksisterede et effektivt samarbejde mellem Københavns og Aarhus Universiteter om hovedfags-statistikundervisningen. I 1965 initierede Aarhus-gruppen Nordisk Konference om Matematisk Statistik, som skulle gå på omgang mellem de 4 nordiske lande; det var Danmarks tur næste gang (1973)
Der foregik forhandlinger om et nordisk statistisk tidsskrift, og der manglede en dansk organisation til at indgå i dette projekt. Professor G. Rasch ville fylde 70 år 21 september 1971, og mange ville gerne fejre denne respekterede men også kontroversielle pionerskikkelse i moderne dansk statistik. Et fagligt selskab ville give en fin klangbund for dette.

Min fremstilling starter med, hvordan den stiftende bestyrelse håndterede disse ganske varierede udfordringer. Vi skal se, hvordan Selskabet lige fra starten havde et imponerende program af foredrag dækkende ganske bredt i teoretisk og anvendt statistik.
Et omfattende ’græsrods’-initiativ i 1979 førte til detaljerede rapporter om spørgsmålet, om Selskabet var for elitært; et udvalgsarbejde foreslog at erstatte ’teoretisk’ med ’anvendt og teoretisk’ i Selskabets navn. Sagen kulminerede på generalforsamlingen 26/2-1980, som vedtog en modererende kurs.
Den centrale aktivitet har altid været møder med foredrag; den mest langtidsholdbare version de såkaldte todages-møder, normalt tirsdag eftermiddag til onsdag middag.
Der har været en lang række samarbejdsmøder med nærliggende selskaber.
De nordiske initiativer (Nordisk Konference om Matematisk Statistik; Scandinavian Journal of Statistics) har vist sig levedygtige.
Selskabet vedtog på generalforsamlingen 26/2-2019 at optage netværket Young Statisticians Denmark som ungdomsafdeling.

Slides og handouts fra foredraget kan findes på:

Mixed membership models for mutational signatures in cancer genomics

Asger Hobolth (AU)

Somatic mutations in cancer can be described as a mixture of different mutational signatures. The signatures can often be attributed to exposures such as UV-light or tobacco smoking, and they can be decoded using a mixed membership model called non-negative matrix factorization. I will describe how non-negative matrix factorization works and why it is useful for cancer treatment and precision medicine. I will then proceed by explaining some fundamental methodological problems with applications of non-negative matrix factorization, and discuss novel solutions developed in my group.

The talk is based on joint work with Ragnhild Laursen (Department of Mathematics, Aarhus University), Marta Pelizzola (University of Veterinary Medicine Vienna) and Lasse Maretty (Department of Molecular Medicine, Aarhus University).

Young Statisticians Denmark presents: Young statistical research in Denmark

Ann-Sophie Buchardt (Section of Biostatistics, UCPH), Sneha Das (DTU Compute) and Nikolaj Thams (CoCaLa, UCPH)

What kind of questions and topics occupy the minds of young statistical researchers in Denmark? This presentation will give you a short tour around a few of our statistical research departments, as three young researchers from Danish universities give a brief introduction to their research. In this way you will get a glimpse of what kind of projects new statistical researchers spend their time on.

Learning to reflect - data-driven strategies for stochastic control

Claudia Strauch (AU)

One of the fundamental assumptions in stochastic control of continuous-time processes is that the dynamics of the underlying process are known. This is, however, usually obviously not fulfilled in practice. On the other hand, a rich theory for nonparametric estimation of the characteristics of continuous-time processes has been developed over the last decades. In this talk, we discuss how to bring together these two areas for developing purely data-driven strategies for stochastic control, which we explore for ergodic singular control problems associated to continuous diffusions and Lévy processes. Applications can be found in many areas of life such as natural resource management, engineering or in the financial area.

Analysis of recurrent events data

Philip Hougaard (Lundbeck/SDU)

Recurrent events data refer to events that over time can occur several times for each individual. The frame for this presentation is events that can occur at most a few times during a clinical trial, such as hospitalizations or heart failures. A simple approach is to analyze such data is by means of a Poisson process model, but it should be clear that this approach ignores the dependence between events, whether created by random subject differences or direct dependence between the events. In clinical trials, a classical and common solution is to consider only the time to the first event and thus avoiding the need to consider potential dependence. This ignores information that is important to the patients and therefore we should search for more complete analysis techniques. There are indeed many suggestions on how to address the dependence in recurrent events data. Common to many of these suggestions is that they are very simple to fit. This simplicity has implied that the area is crowded with techniques, which unfortunately give markedly different estimates for the treatment effect. I will try to make what could be called a 360° review of some of these techniques, following the principle laid out in the classical quote of George Box “All models are wrong, but some models are useful”. By being useful, I particularly focus on whether a treatment difference can be assessed in a fair and relevant way. This is in line with the recent development in the pharmaceutical industry covering estimands, which popularly can be described as an approach to define a treatment effect, covering the real world, where treatment may be stopped due to adverse events, additional treatment may be applied or the patient may die. At the same time, the approach aims at defining the treatment effect directly in the value of variables, instead of relying on assumptions specified in statistical models. This estimand framework is a step forward but may be too formalistic for more complex data types, such as recurrent events in the presence of mortality.