Methods

(1) Archive AEO and (perhaps) other versions of NEMS. In later years, assess forecast errors by running the archived model, replacing assumed values with actual values for important parameters. Construct the runs using an appropriate experimental design to isolate important influences.

(2) For each version of NEMS to be evaluated later, design and run a set of scenarios configured to isolate important influences. Based on these data, construct approximations of important NEMS supply and demand components using regression analyses. Archive the solutions and regression results. In later years, assess forecast errors using the regression equations by differentially replacing assumed with actual values for important influences.

(3) Proceed as in (2) using the standard solution sets prepared for the AEO or special study.

For (1): the size and complexity of the NEMS model and the associated hardware and software development platforms make it impractical to maintain and run non-current versions of the model over extended periods of time.

For (3): The NEMS scenarios prepared for the AEO, or other studies, do not have sufficient variation in assumptions to identify an appropriate roster of individual influences for the purposes of this project.

As a result, method (2) has been adopted for this project.