一类小样本的统计方法建模及其可视化  

A Small-Sample Modeling and Visualization on Statistical Methods

作　　者：游文杰[1,2] 吉国力[1] 袁明顺[1,2] 冯清[2] 

机构地区：[1]厦门大学信息科学与技术学院,福建厦门361005 [2]福建师范大学福清分校,福建福清350300

出　　处：《数学的实践与认识》2013年第7期68-75,共8页Mathematics in Practice and Theory

基　　金：高校博士点专项科研基金(20070384003);福建省教育厅科技项目(JB08244)

摘　　要：针对一类高维小样本数据,利用统计方法的非参数检验与偏最小二乘回归(PLS)构造小样本预测模型,实现基于Wilcoxon秩和检验的变量选择与基于PLS的变量压缩降维.并通过DNA序列分类问题实现基于统计方法的小样本数据建模与可视化,计算结果表明方法对小样本具有可行性、有效性.With a class of high-dimensional ＆ sample are constructed using Hypothesis Test statistical methods,which carried out variables small-sample data, Prediction models of small- and Partial Least-Squares Regression（PLS） of selection on wilcoxon rank-sum test and variables compression ＆ dimension reduction on PLS. And small-sample modeling and visualization on statistical methods are achieved by the instance of DNA sequence classification, the results show that the modeling methods of small-samples have feasibility and stability.

关 键 词：小样本数据 Wilcoxon检验 偏最小二乘回归 DNA序列分类 

分 类 号：O212.1[理学—概率论与数理统计]