Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 57
Procedures and Measures for Further Research 57 4.2.10 S-1: Sample Sizes Sample size is probably the single most controversial item in household travel surveys and one on which there is virtually no agreement, as evidenced by samples ranging from a few hundred to as much as 20,000 households. Even though there have been a number of documents providing guidance on sampling (TMIP, 1996b; Smith, 1979; Stopher, 1982), there seems to be either igno- rance of the existence of these documents or the guidance that they suggest are not accepted. It was hoped to develop minimum sample sizes, based on the purpose of the personal travel survey (model estimation, model updating, regional description, and policy testing and formulation), that would be different from previous guidance, which either offered formulas for calculating minimum samples or provided some possible default values to use in sample-size calculations. Procedures to develop appropriate sample sizes are not lacking either in the transportation field or in the survey sampling literature. Clearly, the fact that there is such a wide variation in chosen sample sizes for household travel surveys arises from at least two issues: (1) available budget and (2) political rather than statistical justification of a particular sample size. Costs for household travel surveys are large compared with any other planning activity. Many smaller MPOs will undertake a household travel survey because the staff feels it is essential, but the sample size will be dictated by available funds. This often leads to a decision to collect data with an inadequate sample because it is felt to be a better option to collect less than the optimal amount of data than to collect no data at all. Furthermore, even though an inadequate sample size may result in modeling problems, models will still be built with what data are available, and too rarely are problems with the models and their forecasts correctly attributed to lack of sufficient data in the first place. It is very possible that no amount of effort in defining adequate or mini- mum sample sizes will ever completely change this situation. Political issues may range from multiple jurisdictional contributions to the survey costs, result- ing in pressures for the sample to be large enough for each contributing jurisdiction to obtain reli- able results to a belief that neither politicians nor the public will accept that a statistically adequate sample will actually be sufficient for the purposes of the survey. An example of both of these issues arose in Southern California in 1990. A statistically adequate sample of the region would be in the range of 3,500 to 5,000 households. However, because funds were being derived from various coun- ties and other jurisdictions in the region, it was essential that each of those jurisdictions receive suf- ficient sample to be able to conduct independent analyses and, in some cases, modeling. At the same time, it was felt that people in the region would not accept that adequate information could be provided for a region with a population of 12 million from a sample of 5,000 or fewer house- holds. The end result was a decision to draw a political sample of about 15,000 households rather than a statistical sample of 3,500 to 5,000 households. Notwithstanding that such situations will arise, it still seems reasonable to specify standardized procedures in sample design that are based on statistical requirements rather than unknown polit- ical requirements. To proceed with this task, it will be necessary to take into account the issues of stratification, error levels, and augment samples and develop simple guidance for sample size from this. Sample sizes should be examined from recent surveys--particularly those that have been used for model estimation, model updating, and policy testing and formulation--and a determi- nation made of the adequacy of the sample for these purposes. Again, we note here that the 15,000 household sample in Southern California turned out to be less than adequate for mode-choice modeling in that region because there were no augment samples and the decision on how to strat- ify the sample resulted in very few transit trips in the final data set--too few, in fact, to allow reli- able mode-choice models to be built with the intended specifications. One of the important issues to consider in setting the sample-size standards is to deviate from previous guidance and not tie the sample size to regional size, except in very broad terms. The