identify relevant studies by a systematic search and then use effect size standardizing on each study result. In addition, this approach corrects sampling error and other artifacts, and it can present an estimate of the total effect with minimized subjectivity (5). Further, since different studies might be influenced by methodologies, time span, and regions, hierarchy meta-analysis has been applied using these factors as grouping variables. Separate analyses are conducted for each group.The process used in the meta-analysis is outlined in Figure 1. After careful review of the included articles, two separate meta-analyses were conducted separately for comparison studies and regression studies.For regression studies, measures of weather conditions were taken as the continuous variable. Effect size and percent change in crash rate were both applied to standardize the research results. However, due to space limitations of this paper, those results are not presented; they may be found in Qiu (6).For the comparison studies, measures of weather conditions were considered as a binary variable (such as snow versus nonsnow). In addition to the overall meta-analysis carried out for each weather category (snow, rain, etc.), hierarchy meta-analyses were conducted separately for the comparison studies stratified by decades (time span) and by countries.
METHOD
Literature SearchA literature search for relevant studies published from 1970 to 2005 was conducted for both peer-reviewed literature and unpublished technical reports and theses. Search strategies used ensured this study contained enough primary studies for meta-analysis, because meta-analysis based on a large number of studies with a small sample size has been shown to be more accurate than that based on a small number of studies with large sample sizes (5).Following the search, 376 papers and reports were selected for further examination, of which 108 were determined to be pertinent. From those reports, 34 that provided 78 result records were selected for meta-analysis.
Study Inclusion Criteria and CodingPrevious studies that explored the association between weather conditions and traffic safety have investigated diverse variables. These studies have also applied a variety of methods and were based on Adverse weather obviously has an impact on vehicle crash rates on roads and highways. However, it would be valuable to quantify the extent to which weather conditions affect the crash rate. To do that, a meta-analysis has been conducted to generalize research findings on this subject and attempt to quantify the impact of weather on traffic crashes. Studies between 1967 and 2005 that examined the interaction of weather and traffic safety were reviewed. Thirty-four papers and 78 records that meet the predetermined criteria were included in the analysis. Crash rates from each study were normalized with respect to effect size for meta-analysis generalization. Results indicate that the crash rate usually increases during precipitation. Snow has a greater effect than rain does ...