1. The reliability of polynomial trend models is highest at the edges of the data (first and last time points) and falls off slowly as we extrapolate.
2. An auto regression is a regression in which the explanatory variables are lagged variables formed from the response.
3. Regression models for time series should use either time trends or lags of Yt as predictors, but never mix the two.
4. The Durbin-Watson statistic can be computed from the autocorrelation of the residuals from the regression.
5. The Durbin-Watson statistic should not be relied upon to evaluate regression models that use lagged values of Yt as explanatory variables.
6. We can only forecast a first-order auto regression one period beyond the end of the observed data because the value for Yn +1 needed to predict Yn +2 is not known.