A Bayesian copula framework to infer ontogenetic relationships and predict associated parameters in human growth and development

Abstract Phenotypes are multidimensional. Yet many contemporary methodologies lack the capability to fully quantify the biological complexity of dynamic and multivariate processes. To address this, we fit a Bayesian Gaussian copula model with a diverse, high dimensional human growth and development dataset using traits related to growth of individual long bones, development and ossification of the skeleton, and development of the dentition. The results demonstrate a modular pattern of human growth and development with the strongest relationships between traits of similar ontogenetic origin. Relationships between traits are hierarchically organized within and between modules and can be further dissected into units related to growth direction, ontogenetic timing of skeletal fusion and dental development. We show that the results above can be translated to practical applications in biological anthropology such as the estimation of an unknown independent variable (e.g. forensic age estimation) or the estimation of a missing response variable (e.g. data imputation). Our understanding of the multivariate human growth and development phenotype is enhanced by this research, and it demonstrates the necessity of appropriate multivariate modelling in both inference and prediction tasks.