目前仍难以有一种具体的函数关系来描述随时间或随地层厚度变化的地质数据,LOWESS方法被证明是拟合这类数据的较为有效的方法。采用LOWESS方法对我国上扬子地区晚古生代98个海相碳酸盐的锶同位素数据进行科学的拟合,使拟合得出的曲线具有更好的全球可对比性,从而提高我国锶同位素地层学的研究水平,并使之与相应的国际研究接轨。同时,还从数学角度讨论参数回归和非参数回归的区别,并对影响拟合结果的两个参数和误差范围进行讨论。
At present,it is difficult to find a specific function to describe the variation of geological data with time and stratigraphic thickness. However,LOWESS method proved to be such an effective method in fitting these data. Based on the mathematical method, 98 data of marine strontium isotope data from the Upper Yangtze platform carbonate in China were selected for the LOWESS fitting. The fitting curves obtained can be compared with the global date of 87Sr/86Srage database and can be used to demarcate the stratigraphic age. Meanwhile, the difference between parametric and nonparametric regression from the view of the mathematics, and the error affecting the fitting outcome were also discussed in the paper.