STATISTICAL-MODELS FOR WORK TRIP RATES OF KUWAITI HOUSEHOLDS

Most trip generation analysis procedures utilized in transport plannin g studies are based on either zonal regression or cross-classification analysis. There has been continuous criticisms of the quality of thes e techniques for a variety of reasons. The use of the Generalized Line ar Model (GLM) framework in conducting household-level trip generation analysis has been the subject of some research, but has not been full y explored, and only a few applications have been reported. There are several advantages that the GLM approach can provide, including the va riety of the statistical models that can be investigated and tested wi thin the general framework, the consistency in the selection of the ex planatory variables, and in the way low-frequency observations are han dled. This paper outlines briefly the generalized linear model framewo rk and describes the various statistical model options it provides for the household-level trip generation analysis, which include regressio n, ANOVA and analysis of covariance models with various link functions and assumptions about the underlying distribution of trip observation s in the classification cells. Data on work trips and household charac teristics from Kuwait generated from an extensive home interview surve y is utilized to demonstrate the practical nature of the proposed fram ework and the possibilities it offers. The application uses data of Ku waiti households in three housing types. Work trips are found to be in fluenced by car ownership and the number of adults and children in the household. Some of the interaction effects are found to be significan t.