The Promise and Peril of Randomness

Eric Lofgren

This playlist explores a broad range of topics, from machine learning to regression to the mean impacting effect estimates. In each, there is an element of randomness, a concept that permeates epidemiological methods but is rarely touched on. This may be that outcomes are correlated with each other and errors are *not* random, the use of random sampling to construct ensembles of decision trees, or random variation in patient outcomes creating spurious outbreaks – and spurious control efforts – in hospital inpatients, a small and highly variable population. They invite the reader to think more critically about what we mean by random, how randomness impacts epidemiological inference, and how we may be able to control – or exploit – randomness in our research.

Leave a Reply