|
|
|
Title: REVISED DECISION CRITERIA FOR BEFORE-AND-AFTER ANALYSES
Accession Number: 00474326
Record Type: Component
Availability: Find a library where document is available Abstract: Because better experimental designs utilizing control sites are not always feasible, a simple before-and-after analysis is commonly used to analyze accident rates and other counted events. Treating the number of events counted before some experimental change as a known constant rather than as a random variable is a fundamental conceptual error that falsely inflates the confidence level at which the experimental change can be judged to have had a significant effect. For example, a reduction in the number of accidents observed after some improvement has been implemented may be judged to be statistically significant when, in fact, it is primarily the result of the chance occurrence of an unusually high "before" count, a typical manifestation of the "regression-to-the-mean" phenomenon. By properly treating the initial count as a random variable, at least a portion of this problem is avoided. New tables are developed to provide more appropriate decision criteria for applications of this type.
Supplemental Notes: Publication of this paper sponsored by Committee on Traffic Records and Accident Analysis.
Monograph Title: TRAFFIC ACCIDENT ANALYSIS, ALCOHOL INVOLVEMENT, AND EVALUATION OF HIGHWAY IMPROVEMENTS Monograph Accession #: 00495821
Report/Paper Numbers: HS-039 850
Corporate Authors: Transportation Research Board 500 Fifth Street, NW Authors: Weed, Richard MPagination: pp 8-17
Publication Date: 1986
Serial: ISBN: 0309040620
Media Type: Digital/other
Features: Figures
(6)
; References
(7)
; Tables
(4)
TRT Terms: Old TRIS Terms: Subject Areas: Highways; Safety and Human Factors; I81: Accident Statistics
Files: TRIS, TRB
Created Date: Oct 31 1990 12:00AM
More Articles from this Serial Issue:
|
