An algorithm to the method of curve fitting by the process of least squares
L. Marcus · 1956
The importance of determining an analytical function to fit a set of experimental data has long since been recognized. Efforts have been made to find reliable methods of relating an analytical function to a set of coordinates, and at the same time, to avoid the dangerous loss of significant figures in the computation. It is well known that the classical theory of least squares is one of the best methods for fitting an analytical function to a set of experimental data. However, using this theory, the exact equation to be fitted must be chosen beforehand. Hence, the final form of an equation using the minimum number of terms required can only be obtained by a method requiring trial and error. The method to be outlined in this paper eliminates the trial and error computation in determining approximately the minimum number of terms in the final equation. It also has the advantage of avoiding the solution of an almost singular set of simultaneous equations.