Using Inverted Relative Entropy to Determine the Representativeness of Samples in Mobility Panels

Travel behaviour is influenced by many factors. Changes in travel patterns can typically be derived from panel data. However, non-random attrition may influence the external validity of the results obtained through panel data. Consequently, it may be necessary to recruit additional respondents between waves. But how to determine how many and which respondents to recruit? In this paper we introduce a new method for determining the sample’s allowed deviation from the population, as based on the concept of inverted relative entropy. We used several working examples to show how the method provides a good representation of the complexity of recruiting respondents who possess various (combinations of) characteristics. The paper concludes with a discussion and recommendations for future research.