The uncertain death. Estimating mortality structure by random sampling

Imprecise data produced by commonly applied osteological age-at-death estimations profoundly affect all research on age-dependent mortality in past societies. With uncertain death estimation (UCD), we propose a novel approach to estimating the mortality structure from imprecise data and present a corresponding R package for simple application. Through repeated random sampling of imprecise age-at-death ranges, UCD estimates the mortality structure of a given skeletal sample. We demonstrate the applicability of UCD in a proof-of-principle study on two samples with known age at death (Bass-Mercyhurst Collection and Coimbra Identified Skeletal Collection). Two case studies of German Neolithic skeletal material illustrate UCD's applicability to archeological samples with dissimilar states of preservation. To comparatively quantify the accuracy of UCD, maximum likelihood estimations, Kaplan–Meier survival estimations, and age-category mortality profiles were calculated for all four study samples. UCD outperforms similar existing procedures while incorporating the uncertainty inherent in osteological data. The proof-of-principle study produced significantly more accurate mortality profiles from UCD than from maximum likelihood estimation and Kaplan–Meier survival estimation. Both archeological case studies indicate UCD's ability to provide meaningful new insight into age-dependent mortality in past societies. UCD allows for comparative studies into age-dependent mortality in past societies without requiring a large sample of precise age-at-death estimations. UCD provides an opportunity for fast and simple analysis of mortality structures on a large dataset without neglecting the information contained in the raw data, thereby facilitating a critical study of patterns in age-dependent mortality on a large scale.