Lets assume I want to estimate some leakage current of a transistor in function of the temperature. For this I will take for example a sixth order polynomial. Will I get different results for the estimated leakage current if I do the fitting (using the least squared error method) and evaluation with the temperature in °C and in K (which basically adds a translation by 273.15 on the x scale)?
I did try it and get a (small) difference when I fit in either °C or K, evaluate the polynomial using the corresponding unit and then translate it to compare the results. The difference is small though, so I am not sure if its just a numerical error.