标题:
变量选择方法在多重共线性问题中的应用—基于全国科技投入产出数据的实例The Application of Variable Selection to Multi-Collinearity Problems—Based on the Research and Development Input and Output Data
作者:
安蕾, 贾慧芝
关键字:
变量选择, 多重共线性, 岭回归, PLS回归Variable Selection, Multi-Collinearity, Ridge Regression, Partial Least-Squares regression
期刊名称:
《Statistics and Application》, Vol.4 No.3, 2015-08-27
摘要:
科研投入是提升一国创新能力的前提，但指标之间往往存在较强的多重共线性问题。本文使用岭回归、PLS回归的方法，把我国31个主要的省市自治区分为两类，依次构建R&D投入–产出模型，以期了解我国R&D投入模式。研究结果表明，不同地区受科技投入指标的影响不同，中西部发展地区受政府及企业投入的影响都很显著，而经济较为发达的省市企业的科技创新意识更强。
A prerequisite for the promotion of a nation’s innovation ability is the input of scientific research, but there are always many multi-collinearity problems among the indexes. In order to know the R&D input-output mode, 31 provinces are divided into two parts to set up ridge regression and PLS regression models separately. The research results show that different areas are influenced by different factors. The Midwest is susceptible to the input of the government and companies, while the technological innovation consciousness of the enterprises in the developed area is stronger.