expert : Modeling Without Data Using

The expert package provides tools to create and manipulate empirical statistical models using expert opinion (or judgment). Here, the latter expression refers to a specific body of techniques to elicit the distribution of a random variable when data is scarce or unavailable. Opinions on the quantiles of the distribution are sought from experts in the field and aggregated into a final estimate. The package supports aggregation by means of the Cooke, Mendel–Sheridan and predefined weights models. We do not mean to give a complete introduction to the theory and practice of expert opinion elicitation in this paper. However, for the sake of completeness and to assist the casual reader, the next section summarizes the main ideas and concepts. It should be noted that we are only interested, here, in the mathematical techniques of expert opinion aggregation. Obtaining the opinion from the experts is an entirely different task; see Kadane and Wolfson (1998); Kadane and Winkler (1988); Tversky and Kahneman (1974) for more information. Moreover, we do not discuss behavioral models (see Ouchi, 2004, for an exhaustive review) nor the problems of expert selection, design and conducting of interviews. We refer the interested reader to O’Hagan et al. (2006) and Cooke (1991) for details. Although it is extremely important to carefully examine these considerations if expert opinion is to be useful, we assume that these questions have been solved previously. The package takes the opinion of experts as an input that we take here as available. The other main section presents the features of version 1.0-0 of package expert.