摘要
本文以3个中华蜜蜂Apis cerana cerana种群为模型,测定常用的30个形态变量数据,通过比较逐步判别分析、主成分分析、聚类分析等多种多变量统计方法的分析结果,找出适合划分中华蜜蜂种群、揭示种群间遗传关系的数据分析方法。结果表明:逐步判别分析对于划分中华蜜蜂种群效果最好,并可进行准确率检验;与聚类分析联用有助于反映种群间遗传关系。主成分分析结果的准确度不高,有样点被分入错误种群,且遗传关系也存在偏差。聚类分析的结果与采用的形态变量和距离-分类方法组合的选择有关,不适合划分中华蜜蜂种群。本研究为中华蜜蜂种质资源调查和种群形态遗传分析等科学问题的研究方法提供参考。
Mutivariate statistical analyses of 30 morphometric variables collected from three Apis cerana cerana populations were used to model the genetic relationships between populations.Stepwise discriminant analysis proved a good method for classifying populations and had the advantage of allowing the percentage of correctly classified specimens to be tested using Cross-Validation.Cluster analysis could assist discriminant analysis reveal the relationship between populations.In contrast,factor analysis assigned some samples to the wrong population and gave much more variable estimates of the genetic relationships between populations.The results obtained from cluster analysis were also relatively imprecise,varying with the variables analyzed,genetic distance and clustering method.
出处
《应用昆虫学报》
CAS
CSCD
北大核心
2011年第1期202-206,共5页
Chinese Journal of Applied Entomology
基金
现代农业峰产业技术体系建设专项资金(NYCYTX-43)
公益性行业(农业)科研专项(nyzx07-041)
福建省自然科学基金(2010J05043)
关键词
中华蜜蜂
形态学
判别分析
主成分分析
聚类分析
Apis cerana cerana
morphology
discriminate analysis
factor analysis
cluster analysis
