A least squares, curve fitting method is developed which uses an ascending series of cosine and sine terms or a series of Chebyshev polynomials in the approximating function for fitting periodic or aperiodic data and data that are not measured at equal intervals. A moving average is first applied to the data and up to 50 semi-smoothed points are selected. Then this set of data is smoothed in the normal manner. A goodness of fit parameter called the coefficient of determination is described, and a method is given whereby the coefficient of determination is utilized to find the optimum number of terms in the approximating function. The curve-fitting method has been incorporated into a computer program, subroutine CURVE, written in Fortran 2 to suit an IBM 7090. A listing of CURVE is given. (Author)
展开▼
