Estimation with Serially Correlated Errors
In this section, several methods of estimating models with serially correlated errors will be explored. We will use least squares with robust standard errors to estimate regression models with serial correlation in the errors. We also consider the nonlinear least squares estimator of the model and a more general strategy for estimating models with serially correlation. In the appendix to this chapter, you will find some traditional estimators of this model as well.
As is the case with heteroskedastic errors, there is a statistically valid way to use least squares when your data are autocorrelated. In this case you can use an estimator of standard errors that is robust to both heteroskedasticity and autocorrelation. This estimator is sometimes called HAC, which stands for heteroskedasticity autocorrelated consistent. This and some issues that surround its use are discussed in the next few sections.