A decision framework for nonsymmetric losses

This paper presents a case study of an application of decision analysis to the estimation of the market price of individual townhouse units where there is asymmetry in the loss function. Selling price is estimated by means of a linear model applied to a sample of comparable properties. The estimated market price is then adjusted for asymmetry in the loss function, overestimation being more serious than underestimation in the present case, with optimal adjustment factors derived under LINEX loss and normal prediction errors. A vague prior distribution is assumed and its parameters are revised based upon sample estimates. The model is applied to an actual townhouse development. The methodology is general and applicable to a wide variety of problems of estimation, prediction, forecasting and control.