|•||The overall population of people who will use real-time information as an in-vehicle traveler information system|
|•||The population of people willing to choose alternate routes when they drive|
|•||The population of people who use media information when choosing alternate routes|
|•||The population of people who are highly stressed in various situations|
|•||The population of people who use specific roadside information|
In the surveys conducted in 1993, for each estimate, the population distribution by gender, age, income, private drivers, commercial drivers, and dispatchers also were needed. Because the population was divided into distinct populations, separate stratum statistics for private drivers, commercial drivers, and dispatchers were also calculated. Each stratum was then subdivided into various socioeconomic levels. This reduced the variance of the sample estimate and provided the required subpopulation estimates for the study.
An initial analysis of the survey data may also provide some direct correlations between variables. If the correlation coefficient, r, is close to one, a ratio estimation may be performed in addition to the stratified sampling estimate.
The general issue of how to design motorist information to impact a target audience is closely tied to the type of motorist information available and our understanding of the behavior and decision-making characteristics of the target commuters. Therefore, it is very important to understand the needs of the motorists by receiving feedback from them. Surveys provide a very effective means of collecting information that is pertinent to the design of a motorist information system. Conclusions drawn then become the basis for converting traffic data into motorist information. Furthermore, surveys help to limit the scope of future in-laboratory and on-road experiments and help to identify specific focuses for future studies.
In conducting a survey, sources of bias are usually apparent. It is important to make an attempt to recognize these sources of bias and then try and avoid them. For example, in the Ng et al. ( 1995) surveys, a high number of nonresponders was expected owing to the size of the surveys (7 to 11 pages).