Some Advance Thoughts on the Data Analysis Involved in Configural Polysampling Directed Toward High Performance Estimates

摘要

Using configural polysampling to identify high-performance estimates proceeds in two phases: (1) estimating the best compromise to make at each configuration polysampled, and (2) choosing a generally-defined estimate whose performance comes close to that of this estimated optimum. In the first step, shadow pricing of the criteria for the difference situations involved is helpful. In the second step, whether one uses a regression or selection-by-sectors approach, or combine both, the same shadow prices are useful. This report sketches some of the approaches to the two steps that seem natural. (Author)