Contribution to screening methodology

Publisher Website
Google Scholar
Abstract
The screening method for determining objective approximate values for the sensitivity coefficients of the output of a computer code to the input data is analyzed. An estimator proposed to solve such problems is used to investigate one of the outputs of a simplified model of the Long-Term Energy Analysis Program (LEAP) of the Energy Information Administration (EIA). The results are presented and compared to the true values of the sensitivity coefficients. The estimator produces results that appear more arbitrary than approximate. The screening problem is analyzed from the point of view of a theory of logical inference, and it is argued that no objective, approximate, and useful solution exists for a code treated as a black box (that is to say, that no assumptions, prior or otherwise, are made about the code, except the ability to run) since no degree of confidence can be placed on the solution. Requirements on additional information needed to produce a solution that could qualify as objective, approximate, and useful are investigated. It is shown that such information can in principle be obtained, and the problem then admits a solution. The emphasis is on the conceptual aspects rather than on the mathematical details of themore » problem and its solution. 2 figures, 4 tables.« less

This publication has 0 references indexed in Scilit: