You confuse the value of function with its parameters.
The matching (fit) measures apply to any equations or models. It is always a function (more or less complex) of the difference in experimental value and calculated from the equation. The consistency of the units of both values is important. But what kind of unit it will be - it does not matter. Regards,
you can also use coefficient of correlation (R2) value instead of NSD for your under consideration model.
For your case, you are little bit messing up the parameters with the units, if you provide any more detail about your problem, then it will be more feasible for any member to solve your problem.