主成分分析
PCA | 根据荷载大小进行筛选,可结合Norm值与指标间相关性
Selecting according to loadings, also can be used in conjunction with the correlationship and Norm value | 可降维以减少变量,体现原始变量 信息
Reducing dimensionality to reduce variables and reflecting original information | 因子载荷的符号有正负性,综合评价函数意义不够明确
Unclear signs of factor loads, unclear meaning of comprehensive evaluation function | [25-27] |
聚类分析
CA | 通过R型聚类,将评价指标分类
Classifying indicators through R-type cluster | 直观,结论形式简明
Intuitive and concise conclusion | 评价指标较多时,不易获得结果
Hard to get result
with numberous indicators | [28] |
主成分-逐步回归分析
PC-SRA | 将PCA筛选的指标引入回归分析,通过显著性检验进行筛选
Regression analysis with indicators screened by PCA, selecting by significance test | 可保留影响最显著的指标,预测精度较高
Retaining indicators with the most significant impact, high prediction accuracy | 当变量对因变量影响小时,结果不稳定
Hard to get stable result when independent variables have a small influence on the dependent variable | [29] |
典范对应分析
CCA | 将对应分析与多元回归分析相结合,每一步计算均与环境因子进行回归
Combining correspondence analysis and multiple regression analysis, regressioning with environmental factors in each step of calculation | 可将样方、对象与环境因子的排序结果表示在同一排序图上
Displaying the sorting results of plots, objects and environmental factors on the same sorting chart | 多应用于土壤指标对植物群落组成的影响,应用范围较小
Mostly used in the influence of soil indicators on the of plant communities, a small application range | [30-31] |
偏最小二乘回归分析
PLSRA | 通过典型相关分析来筛选自变量,提取偏最小二乘因子
Selecting independent variables through canonical correlation analysis and extracting partial least squares factors | 可提供更合理回归模型,直观体现原始变量信息
providing a more reasonable regression model to directly reflect the original variable information | 指标较少时不适用
Not applicable with few indicators | [32] |
专家经验法
Expert experience | 根据经验和研究区域实际情况进行筛选
Selecting based on experience and actual situation of the research area | 筛选的指标的综合反映性较强
Selecting indicators with comprehensive reflectivity | 主观随意性大,评价结果存在差 异性
Different evaluation results caused by large subjectivity | [33] |