DSTS 50-års jubilæum - Thursday

Back to Program (titles, speakers and times)

 

Statistical Testing under Distributional Shifts and its Application to Causality

Jonas Peters (KU MATH)

We are interested in testing a hypothesis H0 for a target distribution P, when observing data from a different distribution Q and assume that P is related to Q through a known shift. We propose a general testing procedure and prove theoretical guarantees, even when the shift needs to be estimated from data. Testing under distributional shifts has several applications: we argue that it may prove useful in conditional independence testing, reinforcement learning, covariate shift and causal inference.

Estimands: The Rosetta stone of estimation or opening Pandora’s box?

Birgitte Rønn (Zealand Pharma), Mette Krog Josiassen (Lundbeck) og Helle Lynggaard (Novo Nordisk)

Estimands have been around for many years, but not until 2010 where the FDA* commissioned report from the National Research Council on the prevention and treatment of missing data in clinical trials was published, we began discussing this explicitly within the pharma industry.
The introduction, (or actually the requirement 😀) that we use estimands was done in an effort to bring clarity and transparency into what treatment effect is being estimated. At the same time, the introduction has given rise to philosophical discussions regarding what it is we have been estimating in our clinical trials, what we should be estimating in our trials and not least whether what we are actually estimating is the relevant treatment effect.
In the presentation we will briefly introduce the estimand concept and framework and illustrate it with a simple example. In addition, some of the interesting challenges we face as statisticians getting to terms with this concept will be discussed.

Reference:
ICH E9(R1): Addendum on estimands and sensitivity analysis in clinical trials to the guideline on statistical principles for clinical trials. https://database.ich.org/sites/default/files/E9-R1_Step4_Guideline_2019_...

A statistical sleuth in the UK criminal justice system

Therese Graversen (ITU)

I will talk about my experiences as an expert witness in various criminal cases involving complex DNA evidence. One example of a challenge is that to a statistician it is natural to carefully tailor the analysis to the data at hand, but it is not straightforward to adopt this approach in the legal setting where analysis methods are largely approved for pre-specified scenarios and, typically, for use by non-statistician experts.

Statistikkens skabelsesberetning og dens mulige endeligt

Lars Kirdan (SAS Institute)

  • Hvorfor fik vi statistikere?: 1.000 års vestlig mentalitetshistorie på 10 minutter
  • Hvorfor blev statistikerne institutionaliseret?: Baggrunden for etableringen af statslige statistikkontorer
  • Data and the Big Why
  • Hvorfor går det så galt?: Nøgleordet for fremtiden er ’Relevans’

The Joy of Pseudo-Values

Per Kragh Andersen (KU SUND)

Survival analysis is characterized by the need to deal with incomplete observation of the outcome variable, most frequently caused by right-censoring, and several – now standard – inference procedures have been developed to deal with this. Examples include the Kaplan-Meier estimator for the survival function and partial likelihood for estimating regression coefficients in the proportional hazards (Cox) model. During the last decades, methods based on pseudo-values have been studied. Here, the idea is to apply a transformation of the incompletely observed survival data and, thereby, to create a more simple data set for which ‘standard’ techniques (i.e., for complete data) may be applied, e.g., methods using generalized estimating equations.

An advantage of this approach is that it applies quite generally to (marginal) parameters for which no or few other regression methods are directly available (including average time spent in a state of a multi-state model). Another advantage is that it allows the use of a number of graphical techniques, otherwise unavailable in survival analysis. Disadvantages include that the method is not fully efficient and that it, in its simplest form, assumes covariate-independent censoring (though generalizations to deal with this have been developed).

We will review the development in the field since the idea was put forward in a 2003 Biometrika paper. Focus will be on graphical methods but the theoretical properties of the approach will also be touched upon.

Analysis of Spectral Data

Line Clemmensen (DTU)

My talk will discuss multivariate statistics and machine learning methods used to analyse spectral data with applications to bio production data, astrophysics, and multi-spectral images.

Biostatistics in twin studies

Jacob von Bornemann Hjelmborg (SDU)

Studies with twins have been taking place at the Danish Twin Research Center since the 1950’s. Thanks to great many people remarkable results on familial influences have been revealed from this unique resource of representative twin data. The statistical biometric modelling takes form in landmark papers by R.A. Fisher published back in 1918 providing a basis for further biostatistical developments.
Having cancer studies with twins in mind we will in this talk focus on such developments, in particular the role of survival analysis in biometrics which has improved knowledgde on causing factors. The within-pair dependence becomes an esssential part of any modelling, and we will consider modern approaches to this end under various scenarios.
Further, the matched twin pair design has proven useful in inferring association by reducing certain confounding elements, however, causal interpretation is of biostatiscal consideration. We will highligt statistical problems, areas of development and encourage to collaboration as the data is quite extraordinary providing profound insight.

The thrilling journey and benefits of analytics at Vestas

Sven Jesper Knudsen (Vestas)

Once upon a time, a Vestas engineer started collecting all the data from our wind turbines for no particular reason and not knowing he was way ahead of his time. But data has been a part of the Vestas DNA ever since. Vestas is an excellent data & analytics use case, from a profitability perspective, how data science has evolved, and how we progress with data. Our journey continues fast as our customer's sustainable energy targets are aggressive, requiring bold and innovative solutions, where data and analytics play an exciting role.

predict(DSTS, n.ahead=50)

Claus Thorn Ekstrøm (KU SUND)

We need to talk. DSTS is turning 50, and it is time to reflect on the deeper questions in life. Where did we come from and where are we going? Are we having a midlife crisis?

I'd like to take the temperature on our relationship. With each other, with the society, with the field of statistics, and with the larger public. Where would we like to go from here and what is needed to ensure that DSTS will thrive in the coming years?