COLLABORATIVE RESEARCH: Multilevel Modeling Analysis of Cross-Cultural Data

University Of California-Davis, Davis CA

Investigators

Abstract

Social and behavioral scientists recognize that conclusions drawn from research in affluent western settings may not accurately represent people in other settings around the globe. Therefore, they are turning to cross-cultural samples to assess, contextualize, and explain the range of human variability. For example, if they find that individuals in subsistence-oriented horticulturalist societies have different norms of fairness in economic transactions than do members of western societies, they might find that the explanation lies in differences in ecology or group size. Thus cross-cultural comparisons help social scientists to build theories of human and social variability that apply to all human societies. To answer such questions, social scientists increasingly rely on cross-cultural samples, with collaborations among researchers who collect comparable data in different settings around the world. As compared to random samples from a single population, however, these studies typically result in complex hierarchical data structures that require advanced statistical methods, but that are well suited to characterize the range of variation in cross-cultural samples. This award will allow two anthropologists to develop statistical methods for the analysis of cross-cultural data, specifically a compilation of wildlife harvests by hunters in 21 subsistence-oriented societies. The analysis focuses on the ways in which hunting proficiency varies across the lifespan, testing the hypothesis that the extension of the human juvenile period relative to non-human primates is an adaptation that promotes the gradual mastery of the complex foraging strategies that distinguish the human niche. More broadly, this research informs expectations about senescence and the role of experience in the maintenance of skills-based performance among aging adults. The compilation of data includes the outcomes of approximately 20,000 hunting trips by more than 1,000 hunters. Whereas previous analyses of cross-cultural data have often relied on aggregations and averages, this award will allow for the develop of multilevel modeling approaches that account for the complex data structure and the challenging nature of the outcome variable, a mixture of zeroes and continuous positive values that is difficult to analyze via conventional statistical methods. The statistical models in this project will demonstrate the range of variation in age-related patterns across study sites, which has implications for the extent to which individual societies can serve as models or analogues for prehistoric contexts. In addition to a substantive analysis of a uniquely expansive dataset of hunting returns, this research will also result in the introduction of methods that can be adapted to similar analyses of unaggregated cross-cultural data. Given the renewed emphasis on cross-cultural research by social scientists, such methods are increasingly needed for the large, structured datasets that result from compilations and collaborations.