The hedonic price method is well adapted to the calculation of relative prices and the estimation of the quality price relationship for a complex product. The main weakness lies in the use of multiple regression for the evaluation of the coefficients when there is very little data and when the variables are correlated. In this article, various methods, statistical and neuronal, are compared from both the predictive capacity point of view as well as that of the facial validity of the expected results. The neuronal approach is globally more successful than PLS regression but neither of the two methods leads to an acceptable solution to the problems of interpretation of the coefficients (signs and values ) stemming from colinearity.
DESMET, P. (2000). Hedonic Prices and Colinearity: An Empirical Comparison of Statistical and Neuronal Solutions. Fuzzy Economic Review: review of the International Association for Fuzzy-set Management and Economy, pp. 61-76.