A Test Strategy for Discriminating Between Autocorrelation and Misspecification in Regression Analysis
指出Durbin-Watson检验不仅对一阶自相关敏感，也对遗漏变量或函数形式错误等设定错误敏感，并提出一种区分两者的检验策略。
T HE Durbin-Watson (DW) test statistic was designed to detect a first-order autoregressive (AR(1)) process among the disturbances in a regression model and has become so common as to be reported in virtually all regression studies. It is well known, however, that it can have substantial power in detecting a nonzero disturbance mean (i.e., specification error or misspecification) due to omission of relevant variables or use of an incorrect functional form (see, e.g., Johnston, 1972; Godfrey, 1973; Kmenta, 1971; Savin and White, 1978; Thursby, 1979), or a process among the disturbances other than AR(1) (Blattberg, 1973). That is, consider the regression