Click for next page ( 26


The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



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 25
25 Recall that the TTI data included crashes from 1997 through In spite of the differences found in the six variables described 1999 while the 17-22 study included data from 20002001. above, differences between the two data sets were not statis- Thus, it is not surprising that speed limits were found to tically significant for the vast majority of data elements. Based increase between the time of data collection for the TTI and upon this finding, combining the two data sets was deemed 17-22 data sets. Similarly, the average weight of the vehicle fleet acceptable. Note that finding differences not to be statistically increased dramatically during the 1990s. In the early 1990s, the significant does not necessarily imply that the data sets are 5th and 95th percentile passenger vehicle weights were 1,800 similar. Users should use caution whenever using the com- and 4,400 lb respectively. By 2002, the 5th and 95th percentile bined database to examine highway or crash characteristics weights had increased to 2,500 and 5,200 lb, respectively. This that are close to the threshold of statistical significance. dramatic increase in vehicle weight would be expected to cause the average weight of crash vehicles to be higher in 2000 and 3.6 Relational Database 2001 than during the 1997 through 1999 period. Hence, the nearly 200 lb increase in average weight between the TTI and The design of a relational database for the purpose of 17-22 data sets is not unexpected. storage and retrieval of crash data was developed and imple- Careful examination of the two data sets revealed that the mented. In addition to the data collected under this study, differences in the width and height of the object struck the crash database also stored data from NCHRP 17-11 and between the two data sets could be attributed to overrepresen- the FHWA Rollover Study. tations in the number of tall trees impacted in the 17-22 data The crash database design revolved around the Oracle and of wide ditches in the TTI data. Note that the increase in server, which is an object-relational database management the number of trees or the number of ditches was not sufficient system providing an open, comprehensive, and integrated to produce statistically significant differences in the object- approach to data management. The crash database was struck category. However, the number of very tall trees (15 m composed of a data file containing different types of elements or more) in the 17-22 data was sufficient to produce significant (e.g., CASE_NUM, CASE_ID, DEPARTURE ANGLE, etc.). differences in the height of the object struck. Further, a rela- A user process (or a client process) and a server process were tively small number of wide ditches in the TTI data produced used for successful communication between users and the significant differences in the width of the object struck. crash database. Together these two processes enabled users to The number of rollovers in the 17-22 data was found to be run various queries on the database. significantly greater than in the TTI data. As shown in Table 13, Access to the crash database could be obtained by directly 59% of the cases from 17-22 involved vehicle rollover com- issuing SQL commands or through the use of an applica- pared to only 50% for the TTI data. A careful evaluation of tion that contains SQL statements. The Oracle crash database each case in both data sets could not provide any explanation processes the commands and returns results to the users. It is for the magnitude of the difference in rollover frequency. The physically located on a server residing at the Nebraska Trans- only possible explanations for the high rollover rate is that the portation Center of the University of NebraskaLincoln. Cur- 17-22 data also had 47% light-truck involvement compared rently, logging in directly on the host computer is supported, to 38% for TTI data. Although light-truck sales were growing i.e., the computer running the Oracle crash database server is during the 1997 through 2001 time frame, the 9% increase used for database access. The communication pathway is in light-truck involvement is unexpectedly high. Further, even established using the inter-process communication mecha- though light trucks are known to have a higher risk of roll- nisms available on the host computer. Logging in via a two- over, the overrepresentation of light trucks is insufficient to tiered (client-server) connection, where the machine on which explain the full magnitude of the difference in rollover rate. the user is logged in is connected directly to the machine run- The rollover rates for both cars and light trucks were found to ning the Oracle crash database server, and via a three-tiered be significantly higher in the 17-22 data than in the TTI data. connection, where users will connect to the Oracle crash data- The 17-22 data had 50% and 69% rollover rates for cars and base server via network server(s) by using a customized appli- light trucks, respectively, while the comparable numbers cation, are possible but have not been implemented. However, from the TTI data were 44% and 59%. Unfortunately, the remote access to the database is available using Windows fundamental differences in rollover rate could neither be Remote Desktop Connection (password protected). Data ele- eliminated nor explained. ment names and definitions are presented in Appendix D.