A Study on Testing and Improving Nonlinear Evaluation Methods for Academic Journals
In this paper, indicator-based comprehensive evaluation methods for academic journals are divided into two groups: linear and nonlinear evaluation methods. In nonlinear evaluation, sometimes an abnormal phenomenon occurs where the final evaluation score decreases while the value of component indicators increases. Regression adjustment method, a new method for test and improvement, is suggested as a solution for the above abnormality. In regression adjustment method the first step is a multicollinearity test after a regression analysis, where evaluation score is used as dependent variable and evaluation indicators as independent variables; the second step is to judge whether multicollinearity occurs or not. In the case of no multicollinearity, one should give up those indicators with negative regression coefficients, and then renew the previous process if there is multicollinearity, however, one should adopt ridge regression approach for estimation and adjustment. The authors conclude that attention should be paid to hidden defects in nonlinear evaluation methods; prudence is necessary before one decides to give up certain indicators; excessive regression-adjustment circles is not recommended; there may be nonlinear evaluation methods free from the requirements for test and improvement; further study is expected on many issues in journal evaluation methods.
academic journals regression adjustment method principal component analysis topsis grey correlation
Liping YU Yuntao PAN Yishan WU
Business school, Ningbo University Institute of Scientific& Technical Information of China
国际会议
北京
英文
218-225
2011-06-07（万方平台首次上网日期，不代表论文的发表时间）