DSTS 50-års jubilæum - Friday

Back to Program (titles, speakers and times)

 

Assumption lean inference for survival data

Torben Martinussen (KU SUND)

Much statistical inference is based on parametric or semiparametric models. Robust standard error estimators exist that are reliable even if the model is misspecified. However, in that case, it is unclear what estimand is targeted and robust inference may be of little use. The approach taken byvan der Laan and co-workers with their scientific roadmap is to define the estimand directly as a function of the underlying probability measure. This is preferable if it is obvious how to pick the estimand which may not always be the case, however. We take a different approach choosing an estimand that corresponds to well known parameters under certain models. Estimation is developed based on the efficient influence function.

25 years more with point patterns in Euclidean space and beyond

Jesper Møller (AAU)

Since DSTS' 25 years anniversary in 1996 there have been many exciting developments in statistical methodology related to analyzing point patterns in Euclidean and more general metric spaces, including the sphere and linear networks (such as road networks and dendrite networks). This talk focuses on some high lights, omitting technical details and presenting some application examples. In particular we consider 1) Spatial inhomogeneity, 2) Spatial point processes specified by a covariance function, and 3) Perfect simulation and doubly intractable distributions. We conclude with briefly mentioning other topics.

Joint Initiative for Causal Inference - from Novo Nordisk perspective

Kajsa Kvist, Randi Grøn og Henrik Ravn (Novo Nordisk)

Novo Nordisk has a priority of increased and sustainable use of data for evidence generation in the continuum of data originating from randomized controlled trials and observational databases. Several academic research collaborations have been initiated with a focus on enabling causal inference without the protection of randomization and on disease understanding including the underlying mechanistic of an intervention.
The presentation will dive into a project within the Joint Initiative for Causal Inference. The project targets research questions originating from several long-term cardiovascular outcomes trials (CVOTs). Treatment effects beyond the intention-to-treat analysis can be challenging as randomisation may no longer protect against selection bias. Issues like adherence to randomized treatment, withdrawals, and initiation of other treatments after randomization, which may even be unequally distributed in the randomization arms, will essentially make the CVOT be a mixture of randomised and observational data. This calls for causal inference methods which can analyse a randomised trial like a randomised trial.

Locally associated graphical models and mixed convex exponential families

Steffen Lauritzen (KU MATH)

The notion of multivariate total positivity has proved to be useful in finance and psychology but may be too restrictive in other applications. In this paper we propose a concept of local association, where highly connected components in a graphical model are positively associated and study its properties. Our main motivation comes from gene expression data, where graphical models have become a popular exploratory tool. The models are instances of what we term mixed convex exponential families and we show that a mixed dual likelihood estimator has simple exact properties for such families as well as asymptotic properties similar to the maximum likelihood estimator. We further relax the positivity assumption by penalizing negative partial correlations in what we term the positive graphical lasso. Finally, we develop a GOLAZO algorithm based on block-coordinate descent that applies to a number of optimization procedures that arise in the context of graphical models, including the estimation problems described above. We derive results on existence of the optimum for such problems.
This is joint work with Piotr Zwiernik, Toronto. See also arXiv:2008.04688