Tuesday, February 2/3, 3.15 pm (11.15 pm Japan), 2021. Seminar in Statistics.

Inference for a class of non-ergodic non-Gaussian regression
Hiroki Masuda, Faculty of Mathematics, Kyushu University
Abstract: We consider statistical inference for a class of non-ergodic locally stable regression models with parametric trend and scale coefficients, when the process is observed at high frequency and the local stable (activity) index is unknown. A detailed asymptotics of the associated (conditional) stable quasi-likelihood estimator is given. In particular, we show that the asymptotic property of the estimator is affected in an essential way by a sort of nonlinearity of the scale coefficient. Also shown is how we can conduct descriptive model selection in the non-ergodic setup.
Seminar slides
Location: Online via Zoom. Please email Krzysztof Bartoszek for invitation to Zoom meeting.

Tuesday, March 2, 3.15 pm, 2021. Seminar in Statistics.

How fully anonymized and aggregated mobile positioning data can help in monitoring the COVID-19 pandemic
Stefano Maria Iacus, European Commission, Joint Research Centre, Ispra, Italy
Abstract: Due to an unprecedented agreement with the European Mobile Network Operators (MNOs) the Joint Research Centre (JRC) of the European Commission was in charge of collecting and analyze data (from 17 operators and 22 European member states plus Norway) to provide scientific evidence to policy makers in order to face the COVID-19 pandemic. Despite the data being of different granularity in time and space, after normalization and integration, a set of indicators and digital products were made available to modelers in, e.g., epidemiology and economics.
Location: Online via Zoom. Please email Krzysztof Bartoszek for invitation to Zoom meeting.

Tuesday, April 13, 3.15 pm (10.15 pm Japan), 2021. Seminar in Statistics.

Quasi-likelihood analysis for stochastic differential equations: volatility estimation and global jump filters
Nakahiro Yoshida, Graduate School of Mathematical Sciences, University of Tokyo
Abstract: The quasi-likelihood analysis (QLA) is a framework of statistical inference for stochastic processes, featuring the quasi-likelihood random field and the polynomial type large deviation inequality. The QLA enables us to systematically derive limit theorems and tail probability estimates for the associated QLA estimators (quasi-maximum likelihood estimator and quasi-Bayesian estimator) for various dependent models. The first half of the talk will be devoted to an introduction to the QLA for stochastic differential equations. The second half presents recent developments in a filtering problem to estimate volatility from the data contaminated with jumps. A QLA for volatility for a stochastic differential equation with jumps is constructed, based on a "global jump filter" that uses all the increments of the process to decide whether an increment has jumps.
Key words: stochastic differential equation, high frequency data, Le Cam-Hajek theory, Ibragimov-Has'minskii-Kutoyants program, polynomial type large deviation inequality, quasi-maximum likelihood estimator, quasi-Bayesian estimator, L^p-estimates of the error, non-ergodic statistics, asymptotic (mixed) normality.
Location: Online via Zoom. Please email Krzysztof Bartoszek for invitation to Zoom meeting.

Tuesday, April 27, 3.15 pm, 2021. Seminar in Statistics.

Optimizing the allocation of trials to sub-regions in multi-environment crop variety testing
Maryna Prus, Department of Computer and Information Science, Linköping University and Faculty of Mathematics, Otto von Guericke University of Magdeburg
Abstract: New crop varieties are extensively tested in multi-environment trials in order to obtain a solid empirical basis for recommendations to farmers. When the target population of environments is large and heterogeneous, a division into sub-regions is often advantageous. When designing such trials, the question arises how to allocate trials to the different subregions. We consider a solution to this problem assuming a linear mixed model. We propose an analytical approach for computation of optimal designs for best linear unbiased prediction of genotype effects and their pairwise linear contrasts and illustrate the obtained results by a real data example from Indian nation-wide maize variety trials. It is shown that, except in simple cases such as a compound symmetry model, the optimal allocation depends on the variance-covariance structure for genotypic effects nested within sub-regions.
Location: Online via Zoom. Please email Krzysztof Bartoszek for invitation to Zoom meeting.