A generative model for exploring differences in mortality associated with stressor exposure risk in bioarchaeological contexts

Generative models are an underutilized tool in bioarchaeology that make it possible to directly interrogate how age-at-death is influenced by varied risk of exposure to stressors, while accounting for factors which are ordinarily invisible to bioarchaeologists. Further, the visibility of suspected differences within populations at the sorts of sample sizes common to bioarchaeology can also be examined, helping to inform interpretation of findings. In the present study, cohorts of 50, 100, 500, and 1000 individuals aged 0 years were generated. Each individual was assigned a frailty value, and to either high or low risk groups. These cohorts were run through simulation models in which exposure to stressors varied according to risk group and the severity of stressors if exposed. The difference in mean age-at-death between high and low risk group for each run was tested for significance using Welch's t -test. The model results are used to identify potential minimum sample sizes for bioarcheological research at which true differences in age-at-death due to difference in stressor exposure are likely to be visible. Small cohorts (50 individuals) had low likelihood of detecting true risk group differences in age-at-death except when the difference in exposure to stressors or the severity of the stressor was great enough to produce a mean difference in lifespan of >20 years. The probability of observing a true difference in age-at-death between risk groups increased when the difference in stressor exposure and/or the stressor severity increased for all cohorts. Therefore, group-level differences in lifespan may not be identifiable in small archaeological samples except where stress or inequality is high. The low reliability of results from small samples reiterates the needs to carefully examine equifinality in bioarcheological research, as demonstrated through the application of this model to a case study which examined the Late Woodland phase of the Dickson Mounds. This application assessed the three potential hypotheses put forth by Goodman and Armelagos (1988) to establish how likely they may be when sample size is not a limiting factor on visibility of potential difference within populations. • Small sample sizes can produce misleading trends in bioarchaeological research • Severity of stressors can determine visibility of within-population differences • Small differences in risk of exposure to stressors is not visible in small samples • Simulation modelling is a mean of testing explanatory hypotheses in bioarchaeology • Simulation modelling can ameliorate sample size limitations in bioarchaeology