论文部分内容阅读
正交多项式是一个富有伸缩性的曲线。它的优点是随时可知拟合是否满足要求。其原理是有限差分法。应用条件为X必须是等间隔的。利用电子计算机按照编就的程序进行计算尤为迅速。如X间隔不等,可用Grandage法处理(参见Biometrics,14,287～289,1959)。在利用正交多项式求回归方程时,回归系数的检验可以同时进行。由于回归系数之间已不存在相关,因此,某一项如果不显著,只要将它剔除即可。我们的目标是获得一个符合要求的一个正交多项式方程:(?)=b_0+b_1X+b_2X~2+bX_3~3+…为了提高拟合工作效率,首先划一线图,以其阵形作为确定最后方程的参考。例如,有的呈直线,有的呈高次曲线;当可信度(1-α)为0.95时各次项系数不显著,在
The orthogonal polynomial is a scalable curve. Its advantage is to know whether the fitting meets the requirements at any time. The principle is the finite difference method. The application condition is that X must be equally spaced. The use of electronic computers to perform calculations based on programmed procedures is particularly rapid. If the X interval is not equal, the Grandage method can be used (see Biometrics, 14, 287-289, 1959). When using the orthogonal polynomial to find the regression equation, the regression coefficient can be tested simultaneously. Since there is no correlation between the regression coefficients, if an item is not significant, simply remove it. Our goal is to obtain an orthogonal polynomial equation that satisfies requirements: (?)=b_0+b_1X+b_2X~2+bX_3~3+... In order to improve the efficiency of fitting, we first draw a line graph and determine its formation as a The final equation is referenced. For example, some are linear and some are high-order curves; when the reliability (1-α) is 0.95, the coefficient of each sub-item is not significant.