欢迎访问林草资源研究

林业资源管理 ›› 2013›› Issue (5): 55-61.

• 科学研究 • 上一篇    下一篇

加权回归估计中不同权函数的对比分析

曾伟生   

  1. 国家林业局调查规划设计院,北京 100714
  • 收稿日期:2013-01-09 修回日期:2013-04-22 出版日期:2013-10-28 发布日期:2020-11-23
  • 作者简介:曾伟生(1966-),男,湖南涟源人,教授级高工,主要从事森林资源监测和林业数表研制工作。Email:zenweisheng0928@126.com

Comparison of Different Weight Functions in Weighted Regression

ZENG Weisheng   

  1. Academy of Forest Inventory and Planning,State Forestry Administration,Beijing 100714,China
  • Received:2013-01-09 Revised:2013-04-22 Online:2013-10-28 Published:2020-11-23

摘要: 利用南方杉木(Cunninghamia lanceolata)的240株建模样本数据,通过建立一元、二元立木材积和地上生物量回归模型,对特定权函数、通用权函数及其拓展的4个权函数的加权回归估计结果进行了对比分析。结果表明:针对某一组具体建模数据而言,采用特定权函数进行加权回归估计的做法是合适的;采用通用权函数进行加权回归估计,对不同的建模数据,其消除异方差的效果并不完全相同;为了使通用权函数具有更广泛的适应性,建议将1/f(x)2调整为1/f(x)n

关键词: 加权回归, 权函数, 异方差, 材积模型, 生物量模型

Abstract: Using the modeling data of 240 sample trees of Chinese fir(Cunninghamia lanceolata)in southern China,single tree volume and aboveground biomass equations based on one-variable and two-variable respectively were constructed,and the fitting results of weighted regression with special weight function,general weight function,and four expanded weight functions were compared.The results indicated that the weighted regression with special weight function would be a good choice for certain modeling data;the weighted regression with general weight function might have different performance on eliminating heteroscedasticity for different data.To expand the adaptability of general weight function,it is recommended that the function expression 1/f(x)2 be revised to 1/f(x)n.

Key words: weighted regression, weight function, heteroscedasticity, volume model, biomass model

中图分类号: