Workshop report: working with repeated choice data

Publisher:
Edward Elgar Publishing
Publication Type:
Chapter
Citation:
Choice Modelling The State of the Art and the State of Practice, 2013, pp. 91 - 106
Issue Date:
2013-01-01
Full metadata record
Files in This Item:
Filename Description Size
2014DaleyHessEckert.pdfPublished version290.32 kB
Adobe PDF
In recent years the study of data containing multiple responses from each individual has become the approach of preference for choice modelling, mainly due to increasing reliance on data from stated choice surveys. Multiple responses offer analysts the opportunity to distinguish within-respondent from between-respondent heterogeneity, to investigate behaviour by the same respondent under a range of attribute level combinations and to study behaviours that may not exist in current markets; moreover there are often cost savings in this approach compared with alternative data capture procedures. Notwithstanding these advantages, the analysis of data containing multiple responses from individuals presents specific issues. It is necessary to make allowance for the correlation of individuals’ repeated responses arising from their specific unmeasured attributes in order to avoid biases in the inferences drawn from the data. The issues involved have been known for some years and methods have been developed to deal with them, but work continues and two schools of thought can be distinguished, each offering distinct insights. The first school takes the approach of modelling responses at the aggregate (sample) level while making allowances for the repeated choice nature of the data either explicitly or through correction approaches. The second approach is to develop a model for each individual respondent, thus bypassing the issue of separating out individual-specific effects
Please use this identifier to cite or link to this item: