Computer automation of general-to-specific model selection procedures

Over the last three decades, the LSE methodology (see Hendry, 1993, for an overview) has emerged as a leading approach for pursuing econometrics. One of its main tenets is the concept of general-to-specific modelling: Starting from a general dynamic statistical model, which captures the essential characteristics of the underlying data set, standard testing procedures are used to reduce its complexity by eliminating statistically insignificant variables and to check the validity of the reductions in order to ensure the congruency of the model. As the reduction process is inherently iterative, many reduction paths can be considered, which may lead to different terminal specifications. Encompassing is then used to test between these, usually non-nested, specifications, and only models which survive the encompassing step are kept for further consideration. If more than one model survives the "testimation" process, it becomes the new general model, and the specification process is re-applied to it. This paper proposes a computer automation of the general-to-specific model-selection process, which we call PcGets (GEneral-To-Specific). Written in Ox (see Doornik, 1998), it is a package designed for general-to-specific modelling of economic processes. In Monte Carlo experiments, the general-to-specific approach of PcGets recovers the specification of the DGP with a remarkable accuracy. The empirical size and power of the specification found by PcGets are investigated and found to be as one would expect if the DGP were known.(This abstract was borrowed from another version of this item.)