Longley, P.; (1984) Comparing discrete choice models: some housing market examples. In: Pitfield, D., (ed.) Discrete choice models in regional science. (pp. 163-181). Pion Ltd: London, UK.

PDF 12794.pdf Download (1MB)

Abstract

Introduction: Since the mid nineteen seventies there has been strong interest within variolls branches of social science in the adaptation of the discrete choice modeling methodology towards a wide range of research problems. This has required recognition of a wide variety of alternative decision-contexts (Landau et a1. 1982) and behaviour-patterns (Lerman, 1979), and has also raised general issues concerning the variable extent to which individual or subgroup choices may be restricted by spatial and temporal constraints. Further interest has been expressed about the spatial and temporal transferability of alternative discrete choice models (Atherton and Ben-Akiva, 1976: Galbraith and Hensher, 1982). This substantive diversification has been accompanied by a variety of technical and methodological refinements of the multinomiallogit (MNL) and multinomial probit (MNP) models, ranging from new estimation procedures (Hausman and Wise, 1978) to the development of less-restrictive, computationally tractable discrete choice model forms (for example, Williams, 1977: Daly and Zachary, 1978). Faced with both a wider selection of methodological tools and a broader spectrum of substantive enquiry, there exists a clear need for formal comparison procedures which the analyst can call upon to evaluate a given model specification or framework. In this paper, I attempt to review briefly some trends amongst recent housing choice studies which employ discrete choice modeling methods. A new procedure is then presented (Hubert and Golledge, 1981; Halperin et al. 1984) which may be used to compare discrete choice models specified and/or structured in accordance with different a priori hypotheses. It is argued that this method fills a gap between existing discrete choice model comparison-procedures which are inapplicable to 'nonnested' model specifications, that is, to competing discrete choice models which comprise totally different variable specifications and that such procedures can usefully aid selection of the discrete choice model most appropriate to any given decision context.

Download activity - last month

Export as XMLJSONCSV

Download activity - last 12 months

Export as CSVJSONXML

Downloads by country - last 12 months

Export as XMLCSVJSON

Archive Staff Only

View Item