Fitness-gain manifold, Longitudinal distribution data, Minimum Sum of Chi-squared approach, Resource heterogeneity
We propose a new way of constructing dynamic decision-making model and a non likelihood based statistical approach for analyzing a new data type: longitudinal distribution data. This longitudinal data records a trajectory of an animal’s dynamic decision-making when continuously exploiting a relative large, but close environment.The ensemble of all hosts contained in the environment is postulated to constitute a manifold of species-specific fitness-gains at any time point, and traverses through two major distinct phases: abundance vs. scarcity of pristine hosts. As such a manifold provides the relative potentials to all possible hosts available for selection, we construct a phase-dependent dynamic decision-making mechanism in a form of a self-adaptive conditional probabilistic model. We devise a Minimum Sum of Chi-squared approach to simultaneously evaluate individual cognitive capability within the two distinct phases and address the validity of the manifold based dynamic decision-making model on the longitudinal distribution data. We analyze three real data sets of seed beetle Callosobruchus maculatus collected from three experimental designs with different degrees of resource heterogeneity. Our statistical inferences are shown to successfully resolve the behavioral ecology issue of whether animal adaptively employs a dynamic decision-making mechanism in response to gradual environmental changes.