I'm not surprised. In physics/engineering related model fitting, there is a double-edged risk of over-parametrization as well as under-parameterization. The f

I'm not surprised. In physics/engineering related model fitting, there is a double-edged risk of over-parametrization as well as under-parameterization. The former is almost always non-physical - it fits but the parameters are meaningless as predictor, conclusions or tying to ANYTHING physical as a design strategy. The latter fits only certain corners of the data space and if you don't know which ones, you can get yourself in to trouble that way as well. The former can be worse than the latter in terms of meaningful use for engineering.