Combined Analysis of Heteroscedasticity and Correlation of Repeated Observations in SP Data



Combined Analysis of Heteroscedasticity and Correlation of Repeated Observations in SP Data

Authors

CHO H-J, Freelance Researcher and KIM K-S, Napier University, UK

Description

A frequently quoted advantage of SP experiments is that it is possible to collect several observations from each respondent. This makes it possible to collect sufficient data with a relatively small number of respondents and a limited survey budget. This

Abstract

A frequently quoted advantage of SP experiments is that it is possible to collect several observations from each respondent. This makes it possible to collect sufficient data with a relatively small number of respondents and a limited survey budget. This also enables investigation the way a respondent responds to different trade-offs ( e.g. Jones 1989; Pearmain and Kroes 1990). All simple methods for analysing SP choice data require the assumption that each observation is independent. However, this assumption is not strictly valid when several repeated choices are made by each respondent. Consequently, a limitation, known as the √íRepeated Measurement Problem√ď has been identified.

The purpose of this paper is to identify the repeated measurement problem. To do this, we apply a variety of approaches to tackle the repeated measurement problems: re- sampling method; observation specific scaling method and individual specific variation method. If there are observation specific variation and correlation of repeated observations, we suggest a methodology to combine the correlation in repeated observations from each individual and the heteroscedasticity of repeated observations from SP data.

The next section briefly reviews the existing studies about the repeated measurement problems. Section 3 summarises the SP survey and data collection. Section 4 investigates correlation in repeated observations by applying a re-sampling method, Jackknife. Section 5 explores the repeated measurement problems by the observation specific variances method. In section 6, we consider repeated measurement problems in terms of individual specific variations. Finally, section 6 summarises the results and suggests a hrther study based on main findings.

Publisher

Association for European Transport