主成分分析与因子分析法适合科技评价吗?

2018-08-11 11:22:58俞立平刘骏
现代情报 2018年6期

俞立平 刘骏

〔摘 要〕针对主成分分析和因子分析广泛应用于科技评价,但是对评价方法选用缺乏检验问题。本文建立了主成分分析与因子分析评价方法适用性的检验框架与检验体系,从评价前检验、评价中检验、评价后检验3个角度进行检验。评价前检验包括KMO检验、Bartlett检验、正态分布检验;评价中检验主要是评价指标信息损失检验;评价后检验主要包括主成分或公共因子的解释力检验、代表性检验、指标单调性检验和权重合理性检验。并以JCR 2015经济学期刊为例进行了实证分析,研究认为,采用主成分分析和因子分析评价必须进行方法适用性检验;因子分析在信息损失较大时不适用于科技评价;主成分分析并不适合评价对象较多的情况。

〔关键词〕主成分分析;因子分析;方法检验;科技评价;学术期刊

DOI:10.3969/j.issn.1008-0821.2018.06.011

〔中圖分类号〕G302 〔文献标识码〕A 〔文章编号〕1008-0821(2018)06-0073-07

〔Abstract〕Principal component analysis and factor analysis were widely used in science and technology evaluation,however,the selection of evaluation methods was lack of inspection.In this paper,the test framework and test system for the applicability of principal component analysis and factor analysis were established.It had three perspectives:pre evaluation,test in evaluation,and post evaluation test.The pre test includes KMO test,Bartlett test and normal distribution test;Test in evaluation included information loss test of evaluation index;The post test includes the explanatory power test,the representativeness test,the index monotonicity test and the weight rationality test.This paper suggested that the method suitability test should be carried out by principal component analysis and factor analysis;Factor analysis was not generally applicable to science and technology evaluation;Principal component analysis was not suitable for the evaluation with more objects.

〔Key words〕principal component analysis;factor analysis;method test;science and technology evaluation;academic Journal

主成分分析与因子分析是两种性质相近的多属性评价方法,由于两种方法均具有降维和不需要主观赋权的特点,在科技评价中得到了广泛的应用,但是对于评价方法的适用性,目前学术界基本采用KMO检验与Bartlett检验,这也是这两种方法自带的检验方法,很少有从其他角度对这两种方法的适用性进行思考的研究。对这个问题进行深入研究不仅能够从理论上丰富多属性评价理论,对于科技评价实践也具有重要意义,可以减少评价方法的误用,从技术层面保证评价的公平公正。

在科技评价中,主成分分析得到了广泛的应用。……

登录APP查看全文