Modelagem de dados categóricos através do procedimento de Chi-Square Automation Interaction Detection (CHAID) e análise de variância: uma aplicação no setor imobiliário

Abstract

This research works explores the use of two different statistical techniques for the explanation of the price potential housebuilding clients are willing to pay for their homes. Data was obtained through 3000 direct interviews in eleven Brazilian cities. The interviews were performed during building sales fairs held in those cities. The questionnaires were structured with more than 100 questions dealing with the preference of homebuyers in terms of specific housing attributes and the households socio-economic characteristics. The relative importance of these attributes was obtained by preprocessing the data with CHAID- Chi-Square Automatic Interaction Detection, a technique that enables the user to get an insight on which variables to include in generalized linear models. Among these models, the analyses of variance technique was used in order to handle categorical data and the interaction between variables. Results show the importance of taking into consideration interregional differences in price determination and the overwhelming influence of monthly income as the major determinant of desired price.