The use of regression techniques for both hypothesis testing and forecasting is also demonstrated. The paper begins with a discussion of the general case of time series analysis. Problems and procedures for correctly estimating time series regression models are presented. In the next section, the procedure is modified to include lagged values of the dependent or independent variables as explanatory variables. Methods of generating and evaluating forecasts from both lagged and nonlagged models follow. The final section focuses on alternative time dependent processes. The nonautoregression assumption crucial to the time series model is emphasized, both because of its frequent violation and because of its potentially devastating consequences for the substantive social science research inferences. Tables, figures, notes, and 30 references are included. For other volumes in this series, see NCJ 72008, 72010, and 72012-14.