Practical considerations regarding estimatingNewhen generations overlap

Researchers studying species in nature often find it challenging to apply methods based on simplistic models of reality. Here I consider how some real-world complications influence demographic estimates of effective population size ( N e ) when generations overlap. The most widely-used model (by Hill) expresses N e as a function of variance in lifetime reproductive output ( LRO ) of the N 1 members of a newborn cohort. Hill’s model assumes stable age structure and constant population size, in which case mean . In real-world applications, researchers often ask whether unbiased estimates can be obtained under the following conditions: (1) When for empirical data; (2) If cohorts are defined at a later age than newborns; (3) If survival to age at sexual maturity (α) is not random; (4) When some or all null parents (those with LRO =0) are not sampled. Using analytical methods and computer simulations, I show that: (1) Because variance in offspring number is positively correlated with the mean, will be biased using raw data when , but this bias can be overcome by rescaling var( LRO ) to its expected value when . (2) The cohort can be defined at any age ≤α, provided that (a) LRO data cover the full lifespan (e.g., production of newborns by newborns, or production of adults by adults), and (b) survival to age α is random. (3) If juvenile survival is family-correlated, defining cohorts at age α avoids upward bias in that occurs if newborn cohorts are used. (4) Missing some or all null parents has no effect on , provided that data are rescaled to .