NEWEY-WEST COVARIANCE-MATRIX ESTIMATES FOR MODELS WITH GENERATED REGRESSORS
NEWEY-WEST COVARIANCE-MATRIX ESTIMATES FOR MODELS WITH GENERATED REGRESSORS
复制标题
DOI：
10.1080/00036849400000034
复制
发表时间：
1994-06-01
影响因子：
2.2
通讯作者：
MCALEER, M
中科院分区：
文献类型：
--
作者：
SMITH, J;MCALEER, M
检验了Newey和West（1987）异方差和自相关一致协方差矩阵在生成回归量模型中的性能。生成的回归量的存在导致干扰项的协方差矩阵是非球形的，具有非零非对角元素和非恒定对角元素。Newey-West方法可能是计算一致性标准误差的一种简单方法，并可用于各种计量经济学软件程序。因此，检验Newey-West标准误的小样本性能似乎是明智的。然而，从Monte Carlo实验的证据表明，纽维-韦斯特程序的性能并不比（不正确的）两步普通最小二乘（OLS）程序，这一发现是由两个说明性的实证应用程序的支持。
The performance of the Newey and West (1987) heteroscedasticity and autocorrelation consistent covariance matrix for models with generated regressors is examined. The presence of a generated regressor results in the covariance matrix of the disturbance term being non-spherical, with both non-zero off-diagonal and non-constant diagonal elements. The Newey-West procedure is potentially a simple method of calculating consistent standard errors, and is available in a wide range of econometric software programs. For this reason, it would seem to be sensible to examine the small-sample performance of the Newey-West standard errors. However, the evidence from Monte Carlo experiments suggests that the Newey-West procedure performs no better than the (incorrect) two-step ordinary least squares (OLS) procedure, a finding which is supported by two illustrative empirical applications.