Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
14 INNOVATIONS IN TRAVEL DEMAND MODELING, VOLUME 2 The synthetic population generator (SPG) in Compre-hensive Econometric Microsimulator for Daily Activity- Travel Patterns (CEMDAP) uses census tractâblock groupâblock level summary tables as con- trol totals for synthesizing households and individuals from the 2000 5% Public Use Microdata Samples (PUMS) data. Some of the summary tables contain the distribution of a single variable, while other tables describe the joint distribution of multiple variables. These tables are used to construct a full multiway dis- tribution by using a recursive merge procedure and the iterative proportional- fitting procedure. The SPG allows the user to specify the choice of control vari- ables from a wide range of census variables at run time. Currently, for the DallasâFort Worth (DFW) appli- cation, four household- level variables and three individual- level variables, are used as controls. The household- level variables are household type (six cat- egories), household size (seven categories), presence of children (two categories), and age of householder (two categories). The individual- level variables are gender (two categories), race (seven categories), and age (10 categories). All other variables in the PUMS data that are required for the activity travel pattern simulator, but not controlled during the population synthesis, are not directly used. Instead, their values are simu- lated on the basis of a suite of models estimated by using PUMS and other sources of data. SCHOOL AND WORK LOCATIONS The âusualâ school and work locations are modeled at the âtopâ level. Every work location zone is con- sidered as an alternative in the choice set (i.e., to avoid large prediction bias, the work location model is not applied to just a sample of zones). However, home location zone, adjacent location zone, and central business district zones are given higher preference in the utility functions. In addition to modeling the âfixedâ school and work locations at the top level, work- related activity (business meeting, etc.) destina- tion choice models are implemented at the activity stop level. OUT- OF- HOME AND IN- HOME ACTIVITIES CEMDAP application for the DFW area includes 11 out- of- home activity types for adults (work, school, work- related, drop- off at school, pickup from school, joint discretionary activity with children, grocery Brief Description of CEMDAP Comprehensive Econometric Microsimulator for Daily Activity- Travel Patterns Chandra R. Bhat, University of Texas at Austin K to 12, and preschool is made in the lower- level models on the basis of the age and enrollment type of the partic- ular person in the sample. NUMBER OF IN- HOME ACTIVITY PURPOSES In the Portland models, in- home activities are distin- guished on three purposes (workâschool, maintenance, and discretionary), but this distinction is made only for the primary activity of the day and is predicted only when the person has no out- of- home activities. This dis- tinction did not appear to add substantially to the explanatory value of the models. That information, cou- pled with the fact that most survey respondents are reluc- tant to provide much detail about their in- home activities, explains why none of the other models distin- guishes between types of in- home activities. Some of the models predict which people work primarily at home: that provides some substitution between in- home and out- of- home work. It does not, however, handle the phe- nomenon of part- time telecommuting, which is the focus of some transportation demand management policies. As a result, there is some interest in predicting work at home as a separate activity type in the Bay Area model if the data will support it.