Tuesday, September 8, 3.15 pm, 2020. Seminar in Statistics.

Trait evolution on phylogenetic trees: inference with guided Markov processes
Moritz Schauer, Mathematical Sciences, Chalmers University of Technology and University of Gothenburg
Abstract: Phylogenetic analysis of genome data of contemporary individuals from extant species of interest can tell about the structure of the phylogenetic tree, including times of speciation. But the physical and ethological traits of the implied ancestors of extant species remain unknown. To infer historic trait evolution from the observed phenotypic traits of extant individuals, we develop a continuous time multivariate latent trait model that allows for trait interdependence. We can infer the model dynamics and trait interactions using a probabilistic programming approach based on guided Markov processes.
Location: Online via Zoom. Please email Krzysztof Bartoszek for invitation to Zoom meeting.

Tuesday, November 10, 3.15 pm, 2020. Seminar in Statistics.

Balance indices for phylogenetic trees under well known probabilistic models
Tomás Martínez Coronado, Computational Biology and Bioinformatics, Department of Mathematics and Computer Science, University of the Balearic Islands
Abstract: The main motivation behind the quantitative study of phylogenetic tree shapes is the belief that they reflect properties of the evolutionary processes that have derived them. In this seminar we shall expose a little survey on what is known thus far about the first and second moments, as well as the limit distributions, of balance indices such as the Colless, Sackin, Cophenetic and Quartet indices under the well known probabilistic models for phylogenetic trees such as the Yule and the Uniform models.
Seminar slides
Location: Online via Zoom. Please email Krzysztof Bartoszek for invitation to Zoom meeting.

Tuesday, December 1/2, 3.15 pm (11.15 pm Japan) , 2020. Seminar in Statistics.

Multi-scale analysis of lead-lag relationships in high-frequency financial markets
Koike Yuta, Graduate School of Mathematical Sciences, University of Tokyo
Abstract: Lead-lag relationships refer to time-lagged correlations between multiple time series. Recent empirical analysis of lead-lag relationships in high-frequency financial markets suggests that there would be multiple lead-lag relationships between a pair of assets. Motivated by this observation, we propose a new mathematical model having different lead-lag relationships at different time scales. Then, we propose a novel estimation procedure to identify those relationships on a scale-by-scale basis from high-frequency observation data. We apply the proposed method to a quote dataset of the NASDAQ-100 assets and identify two types of lead-lag relationships at different time scales.
This is a joint work with Professor Takaki Hayashi at Keio University.
Seminar slides
Location: Online via Zoom. Please email Krzysztof Bartoszek for invitation to Zoom meeting.