Simon Wood教授 講演会 「Generalized Additive Modelling methods for two issues in health/epidemiology」

日時 : 2023年3月6日
場所 : 東北大学情報科学研究科数学教室【情報数理談話会】
講演者 :Simon Wood 教授（School of Mathematics, University of Edinburgh）
於：東北大学大学院情報科学研究科
要旨: Air pollution and Covid-19 offer two recent public health issues with significant worldwide impact. Statistical regression models constructed in terms of smooth functions of predictor variables - generalized additive models - can be very helpful in analyzing data arising in relation to both. This talk illustrates this with two case studies. The first is the modelling of 40 years worth of spatially references daily particulate air pollution data over the UK, with the aim of producing epidemiologically useful pollution burden estimates at different locations and times: the data contain some 10 million observations, and novel methods were required to deal with this data volume. The second case study concerns inference of the daily number of new Covid-19 infections from the clinical data available. GAM like models are useful, but some method extensions are needed to obtain the best possible reconstructions. Interestingly, in the UK context infections appear to have been in decline some time before each of the full stay-at-home lockdowns, with the timing of lockdown coinciding rather with an immediate rapid increase in deaths, which lag infections by several weeks.