Multifactorial descriptive modelling on irregular data designs

Nontensor-product functions are in common use for interpolation on irregularly scattered, multivariate data designs. These functions are flexible tools for descriptive modelling of multi-factorial systems responses if there is an option of placing the interpolation nodes freely, with respect to the data design. The approach is exemplified with three formally simple nontensor-product functions. Some emphasis is put on smooth modelling with pseudo-cubic splines and on optimization of smoothness. The approach can replace “simplistic” descriptive modelling wherever necessary (e.g., when high-quality descriptive models of subsystems behaviour are needed for integration in real-structure models of complex systems). Practical applications to biosphere research are noted.