Principles for Impactful Data Science and Statistics: Lessons Learned from Covid-19 Forecasting and Beyond

• Activity Number: 121 - The WNAR Outstanding Impact Award and Lecture
• Type: Invited
• Date/Time: Monday, August 8, 2022 : 10:30 AM to 12:20 PM
• Sponsor: WNAR
• Author(s): Bin Yu*
• Companies: University of California, Berkeley
• Keywords: impact; data science; covid prediction; domain knowledge; software

Abstract

Impactful data science or statistical research outside statistics is a professional responsibility of statistics as a field and a necessity for its prosperity and funding. In this talk, we distill principles out of the Yu Group’s experience in Spring 2020, when a 12-person rapid-response team used skills of data science/statistics and beyond to help distribute Covid PPE. These principles are useful in general to ensure that our data science and statistics work be impactful in domain fields and society at large. A data science process towards impact includes tapping into domain knowledge (about epidemiology and medical logistics chains in the covid forecasting project), collecting or curating a relevant data repository, developing models relevant for solving the domain problem (for short-term county-level death forecasting in the US in the covid project), and building an accessible platform for sharing visualization (AI machine website in the covid project). Finally, we emphasize dealing with problems that require rapid response, often resembling agile software development.