Abusus Non Tollit Usum: Standardized Coefficients, Correlations, and R 2 s
暂无分享,去创建一个
In a withering critique of statistical praxis, Gary King (1986) has recently warned modelers off standardized regression coefficients, bivariate correlations, and coefficients of determination (R2s). Without denying that these particular statistics are frequently abused (and without taking issue with any of King's other admonitions), I want to argue here that they are not therefore useless-and, in the process, to sketch their legitimate uses.
[1] H. Theil. Specification Errors and the Estimation of Economic Relationships , 1957 .
[2] Thomas Mayer,et al. Selecting Economic Hypotheses by Goodness of Fit , 1975 .
[3] Christopher H. Achen. Interpreting and Using Regression , 1982 .
[4] James G. MacKinnon,et al. Model specification tests against non-nested alternatives , 1983 .