Date: Sat, 24 Apr 1999 11:28:34 -0700 From: "edward doyle" <ejd2nopsamdnet.att.net> Subject: Re: Opinions on the 1999 Saab 9-3?
KBB wrote in message <7frs2l$rqj$1nopsam3.Belgium.EU.net>... > >> Awards - Saab 9000, 1992-96 >> >> "Safest In-Production Car," lowest passenger vehicle fatality rates (for >> cars in production) Insurance Institute for Highway Safety, Oct. 14,1995 >> report (all Saab 9000) >> "Substantially Better Than Average" for injury losses, 1994 > >Stop using those data's, Johan,they depends ON THE CAR AND THE PROFILE OF >THE DRIVER . The profile of the typical Saab driver is very different from >other car's driver : they are older and more educated ( mean less >accidents!) than the average folks. >Cars typically driven by old folks also look better on those studies than a >BMW for example. >And Porsche look like one of the worse car on the road for safety, again >because of their drivers and their speed, not because the Porsche are >unsafe!!!!!!!!!!!!!! >Also car being driven a lot, such as diesel cars in Europe, are involved in >more accidents, not because diesel fuel car are not as safe!!!!!. ( Note >that until recently, Saab did not have diesel cars available!) A reasonable point. But there are TWO ways to use real-world accident data, one of which suffers from the type of problem you describe, the other does not. The question then becomes, which type of statistic is being used in the surveys referenced above? The two ways I can think of using real-world data are as follows: 1. Using "normalized" accident rates, and/or injury/loss rates. By normalized, I mean that they would take the number of accidents/injuries and divide by the number of that particular type of car on the road, to come up with an accident/injury "rate" per 1000 cars, or whatever. This results in an injury rate as a proportion of all such cars on the road. 2. Simply take the *actual* number of accidents and determine the number of injuries resulting from such accidents. This results in an injury rate as a proportion of number of accidents for such a car - very different from No. 1 above. The point is that technique 1 above will strongly depend on the type of driver, and lead to vast differences in results as between types of car - really reflecting what type of person drives them and not necessarily the cars themselves. Technique 2 above should be much less dependant on type of driver - a 30 mph accident is a 30 mph accident whatever "type" the driver is. So, again, which type of statistic is used in the accident statistics referenced above?