The (mis)match between theoretical concepts and statistical procedures: Mapping host developmental tempo axes using structural equation modeling
Developmental tempo axes (i.e., large-small body size continuum, slow-fast paced continuum, life-style continuum, slow-quick return leaf economics spectrum) are theoretical constructs used to represent underlying biological processes that cause organism’s traits to covary (e.g., organism’s with larger body sizes have lower metabolic rates). Because developmental tempo axes determine how organisms influence and respond to the environment, they are increasingly used to identify hosts that contribute disproportionately more to pathogen transmission (i.e., highly competent hosts). Despite a formal theoretical language on developmental tempo axes, researchers have used a variety of conventional statistical methods to quantify them, namely univariate models, principal component analyses, and exploratory factor analysis. Here, we used structural equation modeling methods to investigate the relative ability of these conventional statistical methods to 1) map formal theoretical language on developmental tempo axes and 2) to measure these latent biological processes. We first developed a meta-model to map formal theoretical language. We used our meta-model to derive a causal model specific to mammalian life history traits (e.g., adult body size, litter size, gestation length, neonate body size). We then compared our causal model’s structure and statistical results to conventional methods’ causal structures and statistical results.
The meta-model highlighted that current theory hypothesizes the existence of two independent processes that influence functional traits. The primary developmental tempo process is associated with somatic effort (i.e., the large-small body size continuum; the slow-quick return leaf economics spectrum). For example, building a larger body requires a longer lifespan. The secondary developmental tempo process influences traits via constraints that operate independent of body size (i.e., the slow-fast paced continuum; the life-style continuum). For example, independent of body size, allocating resources towards reducing mortality can reduce reproductive output. Disease ecologists have largely ignored the secondary process or have conflated the primary process with the secondary process. Comparison of our causal model’s structure to causal structures assumed by conventional statistical methods revealed that the later mis-specify the causal relationships assumed by theory and, therefore, potentially incorrectly estimate developmental tempo axes. These results suggest that conventional methods may produce statistical results that jeopardize disease ecologist’s ability to predict the identity of highly competent hosts.