Analyse de sensibilite categorielle: Nouvelle approche de consensus construit dans un contexte de carence de donnees et de divergence d'opinion en matiere environnementale (French text).

摘要

Making a decision in a context of lack of data and divergence of opinion, as it could be the case in a developing country such as Mexico, is a difficult and complex process. The DELPHI technique, a multicriterion analysis method based on repeated interviews of a panel of experts, has been frequently used to look for consensus in complex situations, and more especially in the environment field. In this study, it is shown that within this method, commonly used parameters as a central trend measure and aimed to inform experts of the results from previous consultations missions, are not well adapted to the characteristic of the experts' satisfaction and omit part of the relevant information enclosed in the noise due to data dispersion.; The purpose of this work is to carry out a methodology of a category-based sensitivity analysis to be able to construct a consensus in this type of context. This sensibility analysis is based on the development of a filter of statistics parameters directed towards the characterization of the Clearly Expressed Satisfactions (CES) of experts. The determination of these CES and their respective Category-based Typical Deviation (CTD), as well as two indicators of convergence of opinion (ICO) and data accuracy (IPE) forms the elements responding to the fixed purpose.; The selected example to validate this method is the one of the salt load in the effluvium of the tanneries of the city of Leon, in Mexico, problem that has been approached trying to integrate the general principles registered in the concept of sustainable development.; The principal results of the analysis realized enable to show that, in this particular study, the median was reliable only in 23% of the cases, which confirm in a certain way the original hypothesis. Moreover, the determination of these types of satisfaction open new possibilities of interpretation, more susceptible to understand the characteristics of the problem treated and deduct consequences to elaborate new scenarios-solutions. The inevitable following technical and economical studies destinated to quantify impacts, what expert panel based multicriterion analysis that integrated civil society are not able to provide, are therefore reduced to their strict minimum. The time dedicated to these projects can therefore be reduced in a significant way, which would give rise to lower cost. Finally this post-optimization makes a step forward toward a built consensus, instead of a commitment obtained trough traditional methods.; As a conclusion, the introduced methodology does not pretend to question the post-optimization techniques such as DELPHY, but draws the attention to the reader to the risk to try to interpret, through a mathematical lens only, results based on a group of experts. The methodology of category-based sensibility analysis that rely on the satisfaction of experts can positively complete and reinforce the traditional techniques.