Jones, JH; Gurven, MD; Price, MH
Low-dimensional parametric models of human mortality are important tools in biodemography and comparative evolutionary studies. These models allow researchers to smooth mortality data drawn from small samples, which tend to yield volatile estimates of age-specific mortality hazards as a result of limited observations in adjacent age-classes, without imposing particular age-specific patterns. The Siler competing-hazards model is a theoretically-ap- pealing and commonly-used parametric model for human mortality. The Siler model contains five parameters and captures the overall bath- tub-shape of human mortality without otherwise making large assumptions about the level of different mortality components. Unfortunately, the model can be quite challenging to fit because of correlations between the different parame- ters. Gurven and Kaplan (2007) compiled data on the age-specific mortality of hunter-gatherers and other small-scale subsistence populations. Because of the small samples, which are charac- teristic of studies in anthropological demography, Gurven and Kaplan fit Siler models to the esti- mated hazards to smooth and extend curves to the full age range. We present a greatly improved optimization algorithm for fitting the Siler model to empirical data and reanalyze the age-specific mortality of the hunter-gatherer groups in the original sample. Focusing in particular on four groups (Ache, Hadza, Hiwi, Tsimane), we find that previous fits under-estimate increases in senes- cent mortality and over-estimate the decay of infant/early-childhood mortality. We synthesize these results and present new mortality stand- ards for research into the mortality of people in small-scale subsistence populations.