By Katya Sumwalt (SDSC Intern) and Kimberly Mann Bruch (West Hub Writer)
Anna Panorska of University of Nevada at Reno (UNR) teamed with UNR colleague Tomasz Kozubowski and Marek Arendarczyk of the University of Wrocław in Poland to research how extreme events must be considered in the future. They published their most recent study, entitled Preparing Students for the Future: Extreme Events and Power Tails, in the December 2022 issue of the Journal of Statistics and Data Science Education.
Panorska’s research focuses on how and why traditional statistical models are not sufficient to model extreme events—such as the atmospheric rivers causing disastrous floods in California. And, she works to find tools to solve this problem by using power tails to more efficiently understand extreme events.
What are power tails? They are a phenomenon in statistics where the distribution of a dataset has more extreme values than would be expected from a normal distribution. An example is when values significantly deviate from the mean.
“We concentrate on the intuitive, rather than rigorous mathematical treatment of models with heavy tails,” Panorska and her colleagues write in their paper. “Our goal is to introduce the instructors to these important models and provide some tools for their identification and exploration. The methods we provide may be incorporated into courses such as probability, mathematical statistics, statistical modeling or regression methods. Our examples come from ecology and census fields.”
Panorska explained that these extreme events can be statistically classified as data that follows power tail laws and as a result of the data following these laws, they fail to follow some of the laws used in the typical statistical models. “In a variety of fields—finance, economics, physics, geophysics, seismology and hydrology—extreme events are bound to occur and they must be considered.”
She said that financial crises, increased intensity of hurricanes as a result of global warming, atmospheric rivers driven floods, and countless other events can be modeled better by utilizing power tail thinking. Her work emphasizes the importance of teaching undergraduate and graduate students this version of statistical application. By promoting power tail thinking, she said that students will be able to recognize power tail distributions and have a greater understanding of how to reduce the impacts of extreme events.