Juniorprofessur für Computational Statistics









Drei Studierende, zwei Frauen und ein Mann, sitzen in einem Kreis vor dem Zentralen Lern- und Studiengebäude und unterhalten sich.










RESEARCH




My research bridges methodological development and real-world applications in the analysis of sequential data driven by latent processes, guided by the principle:

“You should model the process that gives rise to the data, not shoehorn the data into a model you happen to have at hand”.




























  • Method development for latent Markov models





    Method development for (continuous-time) latent Markov models


    These include hidden Markov models, state-space models, Markov-modulated Poisson processes, and their extensions, with a focus on irregularly sampled data and computational efficiency. For example, we advance a unified conceptual framework that integrates these models under a common dependence structure, enabling researchers to systematically select and tailor models based on data characteristics—particularly the nature of time (discrete vs. continuous), state space (discrete vs. continuous), and observation process (informative vs. non-informative). By leveraging recursive inference techniques and the R package LaMa for fast maximum likelihood estimation, our work enables a modular, "Lego-like" approach to model building.












  • Applied research on latent behavioural dynamics





    Applied research on latent behavioural and ecological dynamics


    We use latent-state models to uncover hidden processes underlying movement, activity, health trajectories, and social behaviour across ecology, sports, medicine, and social sciences. For example, we identified three distinct foraging strategies — benthic, pelagic, and night divers — in Galápagos sea lions using multivariate hidden Markov models, uncovering how individual variation in latent behavioural states shapes ecological niches and enhances population resilience under environmental change. In sports analytics, we investigated the hot hand phenomenon in basketball by modelling free throws using a continuous-time state-space model with an Ornstein-Uhlenbeck process to represent a player’s latent form, revealing that success probabilities exhibit small but statistically significant persistence over time.