Models for risk prediction in complex data structures

The project will join several well-developed research topics in the field of mathematical statistics in Estonia, with a main focus on the methodology for risk prediction. The project is motivated by practical problems from several different fields (genetic epidemiology, infectious disease epidemiology, life and non-life insurance). The problems are mostly related to data structures where the standard statistical methods are not applicable. The main reasons include non-typical distribution (especially in non-life insurance), censoring and truncation (the data on COVID-19 epidemic), as well as dependence between observations in the sample (in the large biobank cohorts). The purpose of the project is to develop methodology for risk prediction for data structures that exhibit such problems. The results will be used in consulting the Government of Estonia in the COVID-19 situation, collaborative projects with the Estonian Biobank and the leading insurance companies in Estonia.