Modelled systems are often hard to tweak in a justifiable and effective way. Often changing some parameters makes it so adjusting others results in no change. Simultaneously pressing the brake and accelerator of a car can result in no measurable change of speed. An algorithmic method to find related parameters would help dimensionally reduce and understand models. So far we’ve found success with a tree-spanning algorithm that can group related parameters into smallest-possible groups. But the final aim is constructing an algorithm which approximates relevant parameter combinations, giving modelers a detailed view of how to best fit their data.