23 C.F.R. Appendix A to Part 1340—Sample Design
Title 23 - Highways
Following is a description of a sample design that meets the final survey guidelines and, based upon NHTSA's experience in developing and reviewing such designs, is presented as a reasonably accurate and practical design. Depending on the data available in a State, substitutions in this design can be made without loss of accuracy. This information is intended only as an example of a complying survey design and to provide guidance for States concerning recommended design options. These are not design requirements. It is recommended that State surveys of safety belt use be designed by qualified survey statisticians. I. Sample Design A. Sample population: It is recommended that all controlled intersections or all roadway segments in the State (or in the parts of the State that have not been excluded by the 85 present demographic guideline) be eligible for sampling. B. First Stage: Usually, counties are the best candidates for primary sampling units (PSUs). In large States with differing geographic areas, it is recommended that stratification of PSUs by geographic region be employed prior to PSU selection. Counties should be randomly selected, preferably with probabilities proportional to vehicle miles of travel (VMT) in each county. If VMT is not available by county, PSUs can also be selected with probability proportional to county population. When sampling PSUs, States should ensure that an adequate mix of rural and urban areas are represented. In some cases, urban/rural stratification must be employed prior to PSU selection. In other cases, it may be more practical to perform urban/rural stratification at the second sampling stage. C. Second Stage: Within sampled PSUs, it is recommended that road segments be stratified by road type. For example, a two-strata design might be major roads vs. local roads, a three strata design might be high, medium and low traffic volume roads. The sample should be allocated to these strata by estimated annual VMT in each stratum. The sample of road segments within a stratum should be selected with probability proportional to average daily VMT. When enumerating all local roads is impractical, additional stages of selection can be introduced and alternative sample probabilities can be used. For example, census tracts within counties can be selected with probability proportional to VMT, or, if VMT is not available, proportional to the square root of the population. Next, within each sampled census tract, road segments can be selected. D. Sample Size: The following tables are provided as rough guidelines for determining sample size for estimating belt use with the required level of precision. The numbers are based on results from previous probability-based seat belt surveys. E. Example: To achieve the required level of precision, a State with 100 counties would sample 20 counties at the first stage. At the second stage, assuming an average of 100 road segments in each sampled county, a sample of 23 road segments per county would be selected. The total sample size would be 20×460 observational sites. II. Data Collection A. Exact observation sites, such as the specific intersection on a road segment, should be determined prior to conducting the observations. B. Direction of traffic to be observed should be determined prior to conducting the observations. C. If traffic volume is too heavy to accurately record information, predetermined protocol should exist for selecting which travel lanes to observe. D. Observations should be conducted for a predetermined time period, usually one hour. Time periods should be the same at each site. E. To minimize travel time and distance required to conduct the observations, clustering of sampled sites can be done. Sample sites should be grouped into geographic clusters, with each cluster containing major and local roads. Assignment of sites and times within clusters should be random. F. Two counts should be recorded for all eligible vehicles: 1. Number of front seat outboard occupants. 2. Number of these occupants wearing shoulder belts. III. Estimation A. Observations at each site should be weighted by the site's final probability of selection. B. An estimate of one standard error should be calculated for the estimate of belt use. Using this estimate, 95 percent confidence intervals for the estimate of safety belt use should be calculated.
Title 23: Highways
PART 1340—UNIFORM CRITERIA FOR STATE OBSERVATIONAL SURVEYS OF SEAT BELT USE
Appendix A to Part 1340—Sample Design
Determining First Stage Sample Size------------------------------------------------------------------------ Number of Number of counties in State counties in sample------------------------------------------------------------------------10............................................................ 720............................................................ 1130............................................................ 1340............................................................ 1550............................................................ 1660............................................................ 1770............................................................ 1880............................................................ 1990............................................................ 19100-120....................................................... 20130-170....................................................... 21More than 180................................................. 22------------------------------------------------------------------------
Determining Second Stage Sample Size------------------------------------------------------------------------ Number of road segments Average number of road segments in each sampled county sampled in each sample county------------------------------------------------------------------------50............................................................ 1960............................................................ 2070............................................................ 2180............................................................ 2190............................................................ 22100........................................................... 23200........................................................... 26300........................................................... 27400........................................................... 27500-900....................................................... 28More than 1000................................................ 29------------------------------------------------------------------------
