Abstract
Unobserved heterogeneity was introduced in 1920 as a modifier of individual hazards. The concept was termed frailty in demography to describe variation in individual longevity, and has been incorporated in methods for survival analysis. As the frailest individuals are removed earlier from a heterogeneous group, mean hazards appear to decrease over time – cohort selection – leading to some of the most elusive effects in population sciences. Despite the accumulation of documented fallacies induced by cohort selection, the issue remains largely overlooked. I will expose the ubiquity of the phenomenon and propose a unified framework to infer and compare trait distributions, with examples of current interest in epidemiology, ecology, and evolution: (1) Vaccines appear less effective in high-incidence settings. Are they, really1? (2) What is the intrinsic effect of Wolbachia on mosquito susceptibility to dengue viruses2? (3) As populations of bacteria are exposed to antibiotics, their mortality rates decline due to selection for noninherited resistance. How does this affect the measurement of fitness effects of new mutations? (4) What does cohort selection add to the debate between neutral and niche theories of biodiversity?
References
- Gomes MGM, Gordon SB, Lalloo DG (2016) Clinical trials: the mathematics of falling vaccine efficacy with rising disease incidence. Vaccine 34: 3007-3009.
- King JG, Souto-Maior C, Sartori L, Maciel-de-Freitas R, Gomes MGM (2018) Variation in Wolbachia effects on Aedes mosquitoes is a key determinant of invasiveness and vectorial capacity. Nat Commun 9: 1483.