Why Misfit When You Already Have Chi-Squared?
A modification of χ2, called Misfit, is suggested; it provides a comparative goodness-of-fit criterion. The numerical value of Misfit gives the faction of the experimental Signal that remains "unfitted." After this quantity is defined and the motivation for introducing it is explained, some examples of its use are given.
We all know that an unsatisfactory "fit" can be "improved" by using data of lower and lower quality. The purpose of the present discussion is to suggest an additional quantity, readily calculated, which, unlike χ 2, is free from the unwelcome property of appearing to improve the goodness-of-fit with poorer data. The new quantity much facilitates comparison between a single theory and several measurements of varying quality. A suggest name for the new criterion is Misfit.