001: On an Absolute Criterion for Fitting Frequency Curves.

1. If we set ourselves the problem, in its essence one of frequent occurrence, of finding the arbitrary elements in a function of known form, which best suit a set of actual observations we are met at the outset by an arbitrariness which " pears to invalidate any results we may obtain. In the general problem of fitting a theoretical curve, either to an observed curve, or to an observed series of ordinates, it is, indeed, possible to specify a number of different standards of conformity between the observations and the theoretical curve, which definitely lead to different though mutually approximate results. This mutual approximation, though convenient in practice in that it allows a computer to make a legitimate choice of the method which is arithmetically simplest, is harmful from the theoretical standpoint as tending to obscure the practical discrepancies, and the theoretical indefiniteness which actually exist.