A hierarchical framework for correcting under-reporting in count data (seminar)
When: Monday, April 14, 2025, from 2.00 to 3.00 p.m. (CET)
Where: Room 34 of the CU002 building (4th floor - Department of Statistical Sciences), Main Campus (P.le Aldo Moro, 5) Roma
Details of the seminar presentation:
Speaker: Oliver Stoner (University of Glasgow)
Title: A hierarchical framework for correcting under-reporting in count data and its application to UK tornado counts
Abstract: When we collect count data in the real-world, such as counts of disease cases, environmental hazards, or crime incidents, our data can often suffer from under-reporting: this is where the counts that we record are less than or equal to the true count, due to some flawed observation process. For example, events could happen when no one is around to witness them, or some individuals with a health condition may not have access to a doctor to report their symptoms to. To gain a more complete understanding, we can try to correct the under-reporting using a statistical approach and predict the true counts. In this seminar, I will present my first research topic, correcting under-reporting of tornadoes in the UK. I began this work as an undergraduate student and eventually published the method in “A hierarchical framework for correcting under-reporting in count data”. The framework combines a Poisson model for the unknown true counts and a Binomial model for the under-reporting mechanism. Using the framework, we can estimate covariate relationships with the severity of under-reporting and predict the true counts. I will show how the framework can be used to correct tornado counts where under-reporting is expected in areas of low population density. I will also share some reflections from my early career as an academic research and, recently, as a Consultant Statistician for the Food and Agriculture Organization.
This seminar is specifically tailored for Master's students. For further information, please contact <sma-dss@uniroma1.it>