CHAPTER VI – Estimating Asymptotic Covariance Matrices
暂无分享,去创建一个
This chapter discusses the estimation of asymptotic covariance matrices. V n is the average of the variances of Z' t ɛ t plus a term that takes into account the covariances between Z' t ɛ t and Z' t–t ɛ t–t for all t and τ. It is sometimes possible to express variances or covariances of ɛ th as functions of the instrumental variable candidates W th . The results for stationary mixing sequences follow as corollaries to the results for general mixing sequences. White and Domowitz have discussed an heuristically appealing way of choosing l ; however, there is no evidence as yet to demonstrate that any way of choosing l yields an estimator of V, which is a useful approximation in samples of the size typically available to economists.
[1] R. Engle. Autoregressive conditional heteroscedasticity with estimates of the variance of United Kingdom inflation , 1982 .
[2] D. McLeish. A Maximal Inequality and Dependent Strong Laws , 1975 .
[3] H. White,et al. Instrumental Variables Regression with Independent Observations , 1982 .
[4] H. White,et al. Nonlinear Regression with Dependent Observations , 1984 .