The exact moments of ols in dynamic regression models with non-normal errors
The exact moments of ols in dynamic regression models with non-normal errors
复制标题
具有非正态误差的动态回归模型中 ols 的精确矩
DOI:
10.1016/0304-4076(89)90086-9
复制
发表时间:
1989
影响因子:
6.3
通讯作者:
T. A. Peters
中科院分区:
文献类型:
--
作者:
T. A. Peters
This paper studies the finite-sample sensitivity of OLS estimators of the autoregressive coefficient in a first-order stochastic difference equation with i.i.d. errors drawn from an Edgeworth–Gram–Charlier population. Analytic formula for the first and second moments of the estimator are derived using the results of Davis (1976) and Sawa (1972). Also, numerical evaluation of these general formulae is undertaken and the results are presented under selected alternative parameter scenarios. The results suggest that the exact bias is relatively insensitive to skewness and kurtosis in the underlying error distribution but that the exact MSE is relatively sensitive to both skewness and kurtosis and this sensitivity increases as the signal to noise ratio increases.