汽車顏色與碰撞風(fēng)險(xiǎn)汽車安全外文文獻(xiàn)翻譯中英文

1、汽車顏色是否會(huì)影響碰撞風(fēng)險(xiǎn)外文文獻(xiàn)翻譯中英文 （節(jié)選重點(diǎn)翻譯）英文Does vehicle colour influence crash risk?Stuart Newstead， Angelo D EliaAbstractAn often asked question regarding vehicle safety is whether vehicle colour has an influence on crash risk and if so, what is the differential in risk between the various colours of vehicle

2、s available. The major objective of this study was to assessthe relationship between vehicle colour and crash risk through the analysis of real crash outcomes described in Police reported crash data. The study employed induced exposure methods utilising single vehicle crashes as the comparison crash

3、 type. It estimated the crash risk associated with each vehicle colour relative to a reference colour which was chosen to be white. Results of the analysis identified a clear statistically significant relationship between vehicle colour and crash risk. Compared to white vehicles, a number of colours

4、 were associated with higher crash risk. These colours were black, blue, grey, green, red and silver. The association between vehicle colour and crash risk was strongest during daylight hours where relative crash risks were higher for the colours listed compared to white by up to around 10%. No colo

5、ur had a statistically significantly lower crash risk than white although crash risks for a number of other colours were not statistically significantly different from white.Keywords： Crash risk， Vehicle colour， Statistical analysis， Induced exposure IntroductionRelatively few studies have investiga

6、ted the question of whether vehicle colour has an influence on crash risk and if so, what is the differential in risk between the various colours of vehicles available. Identifying a measurable difference would enable a more informed choice to be made. The few studies that have been undertaken to in

7、vestigate the relationship have produced conflicting results and in some instances used methodology which may have led to biased results.1One study attempting to address this question using a case -control design (Furness et al., 2003) concluded that silver cars were 50% less likely to be involved i

8、n a crash resulting in serious injury than white cars. However, whilst the case -control design attempted to match for a number of confounding factors, some that are possibly critical to the results obtained were not included. The first of these is vehicle type. The market sector a vehicle is from m

9、ay determine both its colour and crash risk simultaneously. An example of this is commercial vehicles that are predominantly white and might be expected to have a relatively high crash risk due to high exposure. Second, there may be an association between colour choice and driver safety culture or,

10、more broadly, personality traits of the driver. For example, a more conservative, safety conscious, family buyer may choose silver whilst someone with a more sporting pretence to their driving behaviour, and hence higher crash risk, may choose a colour such as black or yellow.In contrast, a prior st

11、udy by Lardelli-Claret et al. (2002) found that light (white and yellow) coloured vehicles were associated with a slightly lower risk of being passively involved in a collision compared with vehicles of other colours. A paired case -control study was used with data from a database of traffic crashes

12、. Only collisions where one of the drivers committed an infraction were included with violators constituting the control group and other drivers forming the case group. Driver, vehicle and environmental variables were also collected and a logistic regression analysis was run to obtain odds ratios ad

13、justed for certain driver and vehicle characteristics. Although this study attempted to account for a large number of confounding factors, perhaps to a greater extent than Furness et al. (2003), the study design still contains the inherent weakness that it is not possible to match controls or to oth

14、erwise adjust the results for unknown, unmeasured or unmeasurable confounding factors.In their paper on the influence of colour on fire vehicle accidents, Solomon and King (1995) use probability theory to show that lime-yellow/white fire vehicles are significantly statistically safer than red and re

15、d/white fire vehicles. They refer to research by Allen (1970) which states that the luminance of a well-worn highway and 2the luminance of its scenery are about the same. Light colours, which include white and yellow, are the most visible against this background. They also point to work by Southall

16、(1961) which states that the relative brightness of the various parts of the daylight spectrum as it appears to a normal bright-adapted eye reaches its peak somewhere in the greenish-yellow region.As evident from the difference in results between the studies previously undertaken, there is still som

17、e uncertainty about the role of vehicle colour in influencing crash risk. Consequently, a need was identified to further investigate the issue, ideally avoiding the use of case -control methodologies and the inherent weaknessespotentially evident in this approach with respect to not being able to co

18、ntrol for factors such as vehicle exposure and driver personality which are difficult to measure. The major objective of this project was to re-assessthe relationship between vehicle colour and crash risk through the analysis of real crash outcomes described in mass crash data. The study aimed to ac

19、hieve this through application of an alternative study design based on induced exposure methods in order to address the apparent weaknesses in prior studies.The generic null hypothesis being tested by the study was that there was no association between vehicle colour and crash risk. This was assesse

20、dagainst the generic alternative hypothesis that there is differential crash risk between at least two vehicle colours, all other factors being held constant. As will be made clear in Section 3, these hypotheses have been tested across the light vehicle fleet as a whole as well as within particular

21、strata of interest. The alternative hypothesis is two-sided since there is no consistent evidence from previous studies to give any a priori expectation about the direction in which various vehicle colours might influence crash risk.DataPolice reported crash data from Australian crashes during 19872

22、004 formed the basis of the data used in this study. In particular, as vehicle colour is only recorded in crash data from the states of Victoria and Western Australia, only data from these states was used. Nonetheless, there was considered to be sufficient crash data to 3achieve meaningful analysis

23、outcomes. In Victoria, the Roads Corporation (VicRoads) maintains a database of all crashes reported to the Police. Crashes are reported to the Police if a person is killed or injured. The Victorian data covered 102,559 drivers of 1982-2004 model light vehicles involved in Police reported crashes du

24、ring 1987 -2004. In Western Australia crashesmust be reported to the Police if anyone involved is killed or injured or the crash results in property damage greater than A$1000 (2001). The Western Australian Department of Main Roads maintains a database of all Police reported crashes. The Western Aus

25、tralian data included 752,699 drivers of light vehicles manufactured between 1982 and 2004 involved in Police reported crashes during 1991 -2004. Out of the 752,699 drivers, 103,544 were killed or injured.For the analysis, the Victorian and Western Australian data were combined and variables which a

26、llowed for stratification along the lines of key variables of interest were prepared. These included light condition at the time of crash, vehicle type, crash injury severity and state. In addition, the relatively large number of variations on vehicle colour in the Western Australian data was reduce

27、d to match the 17 colour definitions used in the Victorian data. The colour classifications used were black, blue, brown, cream, fawn, gold, green, grey, maroon, mauve, orange, pink, purple, red, silver, yellow and white. Clearly, there is potential for significant variation in actual vehicle colour

28、 hues and saturation within each colour category. However, this was the maximum resolution the data allowed for analysis. Cases involving commercial vehicles (a majority of which are white) and Victorian taxis (which are mostly yellow) were discarded from the analysis due to significant differences

29、in the way these vehicles are often used and the narrower range of environments (typically urban) in which they are used. Furthermore, the relative uniformity of colour of these vehicle types did not allow adequate contrasting in the analysis design.DiscussionGovernment agencies and automobile clubs

30、 in Australia report a high level of consumer interest in the issue of crash risk related to vehicle colour. As identified in the literature, there have only been a small number of studies attempting to address 4 this issue. Unfortunately the results published from these studies are conflicting whic

31、h does little for providing solid consumer advice. The two major prior studies attempting to research the issue have both been based on case -control designs. Case -control designs can provide good statistical analysis power when examining rare outcomes, hence their popularity in medical research. H

32、owever, the strength of the analysis outcomes relies heavily on how well the control crashes are matched to the case crashes on all important factors affecting the outcome of interest apart from the factor being assessed. As noted, there are some concerns about how well the cases and controls have b

33、een matched in the previously published studies on vehicle colour and crash risk.Ideally, crash risk would be best assessedthrough the use of an appropriate exposure measure suchas vehicle travel. Such exposure measures aregenerally not available on a vehicle by vehicle basis required for studying t

34、he effects of specific vehicle properties such as colour. Furthermore, exposure based crash risk measures still need to be adjusted for the influence of confounding factors that may differ between levels of the factor of interest before unbiased estimates of the influence of the factor of interest o

35、n crash risk can be assessed. Like the case -control study, the confounding factors are not always available in the analysis data and in some instances cannot be readily measured.Use of an induced exposure design for this study offered the potential to overcome the lack of suitable exposure data whi

36、lst making maximum use of the available crash data. It also offered potential to overcome the limitations of case -control studies and address the research hypothesis in a different analytical setting. Certainly, driver behavioural characteristics associated with vehicle colour choice which are diff

37、icult to measure in the case -control study can be better controlled in the induced exposure setting since vehicle crash distributions within the same colour type are being compared. However, induced exposure studies can have some of the similar failings to case -control studies if the crash type fr

38、om which exposure is induced does not appropriately represent exposure in the group affected by the factor of interest. In other words, althoughthe exp“osinudreucginoges ”a longway to address the weaknessesof case -control studies, the use of such methods cannot remove all possible confounding. For

39、example, the induced exposure crash group includes all single vehicle run-off-road crashes. These crash types are associated with young drivers, alcohol use, speeding and rural roads. Clearly it is possible that the exposure group does not necessarily fully accurately represent exposure in the group

40、 affected by colour, which could lead to biasesin the results. Without access to detailed exposure data it is difficult to quantify this bias.Key confounding factors identified in this study include type of vehicle being driven, the light conditions and jurisdiction of crash. Using the stratified in

41、duced exposure design allowed the adjustment of the analysis results for the confounding influences of each of these factors. Conveniently, it also allowed differential effects of vehicle colour on crash risk to be assessedwithin the levels of each of these confounding factors. The fact that differe

42、ntial relative crash risks by vehicle colour were found between the levels of the confounding factors justifies the need for their control in the analysis. Whether there are other factors that should also have been controlled for in the analysis is difficult to assess and ultimately must be judged o

43、n the intuitive nature of the results and their relationship to other established research data.The results from the analysis using the induced exposure study design, presented in Section 3 show that vehicle colour effects on crash risk were greatest in daylight hours compared to twilight or dark ti

44、mes. Of the statistically significant results for daylight hours, black, grey and silver vehicles were estimated to have the highest crash risks relative to white of 12%, 11% and 10% higher, respectively. The other statistically significant results were for the colours blue and red which were also e

45、stimated to have higher crash risks relative to white. It should be noted that although these colours had a statistically significantly higher estimated daylight crash risk than white, it was not possible to statistically differentiate the relative crash risks of these colours. The estimated relativ

46、e risks for gold and yellow vehicles were close to 1 and based on sufficiently large quantities of data to conclude that these colours do not have a different crash risk to white during daylight hours with a high degree of statistical confidence. For a number of other colours, the results of the stu

47、dy indicated 6a possibly higher crash risk than white but were inconclusive due to the limited quantities of crash data for vehicles of these colours. They included maroon, brown, cream and purple. There was also an indication that orange vehicles may have a lower daylight crash risk than white alth

48、ough this again was inconclusive due to limited crash data quantities. Further analysis based on larger crash data quantities is necessaryto determine the relative crash risks of these colours compared to white. Additional crash data would also allow the statistical differentiation of the relative c

49、rash risks of those vehicle colours with crash risks significantly higher than white.Although relative risks for the twilight and dark analyses were generally less, there were still some statistically significant results. For the dawn or dusk stratum, the results for black and silver vehicles were a

50、gain statistically significant with a nearly statistically significant result being achieved for grey. In this case black vehicles were found to have a 47% higher crash risk relative to white, silver 15% higher and grey 25% higher.The results for silver vehicles in particular are in direct contrast

51、with the study by Furness et al. (2003) which concluded that silver cars were 50% less likely to be involved in a crash resulting in serious injury than white cars. However the results are generally consistent with the investigation of Lardelli-Claret et al. (2002) which found that light (white and

52、yellow) coloured vehicles were associated with a slightly lower risk of being passively involved in a collision compared with vehicles of other colours.It was stated earlier that cases involving commercial vehicles (a majority of which are white) and Victorian taxis (which are mostly yellow) were re

53、moved for this study. The strength of using induced exposure methods lies in the ability of the induced exposure group to take into account influencing factors, especially unknown, unmeasured or unmeasurable confounding factors. In this way the inherent weaknessesof prior vehicle colour studies were

54、 largely overcome. Although the stratified induced exposure methods have the advantage of accounting for confounding factors effectively, the method by which the colour affected and induced exposure crash groups were identified relies on an assumed constant proportional 7exposure split between the c

55、olour dependent and induced exposure crash types across all vehicle colours. For this reason it was deemed sensible to remove commercial vehicles and taxis. Such vehicles are driven predominantly in the metropolitan location and would therefore be more prone to the types of crashes that occur there,

56、 in particular, vehicle to vehicle crashes and crashes where a pedestrian was struck. This would see a bias towards inclusion of these vehicles in the colour affected crash group. Including taxis would see a biased crash risk estimated for yellow vehicles whilst including commercial vehicles would b

57、ias the results for all vehicle colours which were estimated relative to the colour white.One key limitation in interpreting the results of this study relates to the definition and interpretation of the vehicle colour classifications. Obviously, vehicle colours cover an almost infinite spectrum of h

58、ues and saturations. Due to pragmatic requirements, vehicle registers are forced to record these into a fixed number of discrete classifications. Consequently, there will be significant variation in the tomes of colour classified within each of the 17 colour groupings used for this study. For exampl

59、e, blue can range from a shade close to black up to almost silver. With such a variation, the associated crash risk could vary significantly within the one colour classification. To be borne in mind when interpreting the results is that they represent the average risk across all colour variations wi

60、thin the defined category of vehicle colour.A further limitation of the study is that using vehicle colour as the predictor of crash risk does not necessarily give an insight into the physical mechanisms leading to the increased crash risk. The study merely shows and association between certain colo