Analysts typically divide available sales into two groups, one for calibrating the models, known as the learning group, and the other for testing those models, known as the testing group or validation group. The testing group is withheld from the model calibration process so that the performance of the model can be determined based on sales that are not used in the calibration and that the model cannot be fine-tuned to. This process has been criticized as giving up data crucial to the modeling process. The purpose of this paper is not to debate the relative advantages or disadvantages of linearization; rather, the intent is to examine the problems imposed by giving up crucial data in the modeling process and to propose better alternatives.
Valuation - statistical methods, Mass appraisal techniques
This material was first presented at the 15th Annual GIS/CAMA Technologies Conference held February 28–March 3, 2011, in Memphis, Tennessee. The conference is jointly sponsored by the International Association of Assessing Officers (IAAO) and the Urban and Regional Information Systems Association (URISA).
Jensen, D. L. (2011). The use of cross-validation in CAMA modeling to get the most out of sales. Journal of Property Tax Assessment & Administration, 8(3), 19-40. Retrieved from https://researchexchange.iaao.org/jptaa/vol8/iss3/2