Multiday Samples, Parameter Estimation Precision, and Data Collection Costs for Least Squares Regression Trip-Generation Models

Abstract

In this paper it is shown that substantial benefits may be achieved by employing a multiday sample in least squares estimation of linear regression models of travel behaviour. Specifically, by using a multiday sample one can reduce data collection costs while maintaining a given level of precision in the parameter estimates, or one can obtain improved precision in the parameter estimates for a given data collection cost. Furthermore there is an optimal length of observation period that either minimizes the cost of data collection (for a given level of precision in the parameter estimates) or maximizes the precision of the parameter estimates (for a given data collection cost). The characteristics of the optimal multiday sample depend on the cost structure of data collection and the degree of intrapersonal variability in the aspect of travel behavior being modeled. The results also clearly show the superiority of two-day samples relative to traditional single-day samples, under a wide variety of circumstances. Empirical analysis, using data collected in Reading, England, is used to verify the analytic relationships and to assess the benefits of multiday samples in a particular context.