Statistical Bias in Regression and Correlation Studies Due to Uncertainty

 

Nithin Sivadas

Postdoctoral Researcher
Catholic University of America
Space Weather Lab, NASA Goddard Space Flight Center

Wed, March 1, 2023 - 4:00 PM

Uncertainty in the independent variable can lead to unintended biases in correlation and regression studies. Space physicists are interested in characterizing how hot winds from the sun affect the space environment around the Earth. Extreme space weather can lead to spacecraft malfunction and failure of electric power grids, communication networks, and navigation systems. However, since we measure the solar wind far from our planet, we have a certain degree of uncertainty in the solar wind that impacts the Earth. These uncertainties, if unaccounted for, can lead to a misleading perception that the impact of extreme space weather on our planet is lower than it is. In other words, the result of extreme space weather on the planet can be at least 300% higher than we currently believe it to be. In this talk, I will explain how statistical biases appear in correlation studies due to uncertainties in the independent variable and discuss an example from space physics. The underlying statistical effect will interest anyone engaging in correlation or regression studies (including machine learning) in several disciplines.

Refreshments served at 3:45 PM

If you have any questions about the Colloquium Series  or would like to make a donation please contact the Physics Department, cua-physics@cua.edu or (202) 319-5315.