文献类型：期刊文献

中文题名：利用2种非线性混合效应模型(2水平)对杉木林胸径生长量的分析

英文题名：Analysis of the Basal Area for Chinese Fir Plantation Using Two Kinds of Nonlinear Mixed Effects Model(Two Levels)

作者：符利勇[1] 李永慈[2] 李春明[1] 唐守正[1]

第一作者：符利勇

机构：[1]中国林业科学研究院资源信息研究所;[2]北京林业大学理学院

年份：2012

卷号：48

期号：5

起止页码：36-43

中文期刊名：林业科学

外文期刊名：Scientia Silvae Sinicae

收录：CSTPCD;;Scopus;北大核心:【北大核心2011】；CSCD:【CSCD2011_2012】；

基金：林业公益性行业重点项目"我国典型森林类型健康经营关键技术研究"(201004002);国家自然科学基金项目"基于森林清查数据的乔木林碳储量计算方法研究"(31070485)

语种：中文

中文关键词：非线性混合效应模型;2水平非线性混合效应模型;杉木胸高断面积;最佳拟合模型

外文关键词：nonlinear mixed effects model（NLMEM） ; two-level nonlinear mixed effects model; basal area for ChineseFir; the best fitting model

分类号：S758.1

摘要：选择2种2水平非线性混合模型对杉木林胸径生长量进行分析,其中模型1为一般的2水平非线性混合模型,模型2在模型1的基础上进一步考虑固定效应参数随某一特定因子水平变化而变化。本文通过对这2种模型的分析,首先确定构建2水平非线性混合模型的基础模型,然后对模型1衍生出的665种模型及模型2衍生出的2703种模型进行计算和比较:对于模型1,有57种模型计算收敛,当形参b0同时考虑区组和样地效应、而b4和b5只考虑区组效应时,模型拟合效果最好,因此把该模型作为模型1的最佳拟合模型;对于模型2,有24种模型计算收敛,当形参b5同时考虑区组和样地效应、b1只考虑区组效应并且固定效应b0的取值与各区组水平有关时,模型拟合效果最好,因此把该模型作为模型2的最佳拟合模型。最后对传统的非线性回归模型、模型1及模型2进行比较:模型1和模型2的拟合效果都比传统的非线性回归模型好,且模型2的拟合效果最好。
Nonlinear mixed effects model（NLMEM） is based on the relationship between the fixed and random effects in the regression function. The NLMEM has a competitive advantage in analyzing repeated measures data, the longitudinal data and multilevel data. This paper chose two kinds of two-level nonlinear mixed model to analyze basal area growth for Chinese Fir（Cunninghamia lanceolata）. Model 1 is a general two-level NLMEM and model 2 is based on model I to further consider the fixed effects parameters changes with a specific factor. Firstly, through the analysis of these two models, this paper defined the basic model to build the two-level NLMEM. Secondly, 665 kinds of models derived from model 1 and 2 703 kinds of models derived from model 2 were calculated and compared. The results showed that： for model 1, there were 57 kinds of models converging, and when the formal parameter b0 considered the block effects and plot effects, b1 and b4 only considered the block effects, the model fitted the best. For model 2, there were 24 kinds of model converging,and when the formal parameter b5 considered the block effects and plot effects, b： only considered block effects and the fixed effects b0 changed with any level of block level, the model 2 fitted the best. Finally, by comparing the traditional nonlinear regression model, model l and model 2, the results showed that model 1 and model 2 fitted better than the traditional nonlinear regression, and model 2 was best fitting model.