P
Paul Smith
Guest
On Wed, 5 Mar 2003 14:48:05 +0000 (UTC), Tim Woodall <[email protected]> wrote:
>>>>>>> http://www.safespeed.org.uk/fatality.html
>> I've spent hours and hours on the same problems. One way to get noise reduction is to run a
>> rolling average, and I've done that for 2,3,4,5 and 10 years. All it appears to do is confirm the
>> exponential decay curves I'm using. I've seen no better long term fits than exponential decay.
>I've done a least squares regression fit on a 10 year rolling average using an exponential trend
>and the data fits well up until the years including 1999 onwards. But we already know that the
>method of data collection changed in 1999, all of the stats produced have a footnote "Data from
>1999 onwards cannot be directly compared to years before 1999".
The change affects annual total billion vehicle km and amounts to only about 0.2%. We can do little
except neglect it. I wonder if they've done it to distract us? Probably not.
>And I don't know how you can possibly point to a particular year when you have a 10 year rolling
>average. But at the very least, if there really was a step change in 1991, I would expect it to be
>obvious in the rolling average by 1986-1996 but it isn't (at least to me).
Ten years is probably wrong. I was much happier with 3 and 5 years.
In such case you have to compare y-5 with y+5.
The good years in the early 90s were due to the recession; reduced annual mileages and probably
reduced "high living".
>Taking the 10 year improvements,
>1960-1970 through to 1976-1986, accident rates fell by about 40% per decade.
>1977-1987 through to 1981-1991 accident rates fell by about 45% per decade although this isn't as
>consistent a reduction, even when averaged over 10 years.
>1982-1992 thought to 1986-1996 accident rates fell by about 50% per decade
Mainly due to the recession 1990 - 1993.
>1987-1997 through to 1989-1999 accident rates fell by about 45% per decade.
Same again.
>Beyond 1999 the method of stats collection changed but the figures are sitting somewhere abound
>35-45% for the two years I have. Yes the decade 1991 to 2001 is the worst in the period I have
>considered but this is only one period, there were worse periods in the 1960-1970 through
>1976-1986.
>Finally, there were four exceptional years in succession, 1991 - 12% fall, 1992 8% fall, 1993, 10%
>fall, 1994 7% fall. These four years occured after the magic 1991 "speedcamera" year. I would
>expect speed cameras to be placed first at the worst accident blackspots and then only later get
>installed at the less "accident prone" sites so although 1991 may only have had a few speed
>cameras, they were probably sited at the places which could give the best improvement. I remember
>the Bham New Road that regularly used to have serious accidents because of cars jumping red lights.
>You could almost overnight see the difference once the traffic light cameras got installed. There
>were a few speed cameras on this road but, in general, if people drove at speeds where they could
>stop for a red light, then then generally wouldn't be exceeding the speed limit much anyway. (I
>rarely drive this road now and am not local so I don't know whether the accident improvements have
>been maintained but I hope so)
Speed cameras were trialed in 1992 and started to be used seriously in 1993.
--
Paul Smith Scotland, UK http://www.safespeed.org.uk please remove "XYZ" to reply by email speed
cameras cost lives
>>>>>>> http://www.safespeed.org.uk/fatality.html
>> I've spent hours and hours on the same problems. One way to get noise reduction is to run a
>> rolling average, and I've done that for 2,3,4,5 and 10 years. All it appears to do is confirm the
>> exponential decay curves I'm using. I've seen no better long term fits than exponential decay.
>I've done a least squares regression fit on a 10 year rolling average using an exponential trend
>and the data fits well up until the years including 1999 onwards. But we already know that the
>method of data collection changed in 1999, all of the stats produced have a footnote "Data from
>1999 onwards cannot be directly compared to years before 1999".
The change affects annual total billion vehicle km and amounts to only about 0.2%. We can do little
except neglect it. I wonder if they've done it to distract us? Probably not.
>And I don't know how you can possibly point to a particular year when you have a 10 year rolling
>average. But at the very least, if there really was a step change in 1991, I would expect it to be
>obvious in the rolling average by 1986-1996 but it isn't (at least to me).
Ten years is probably wrong. I was much happier with 3 and 5 years.
In such case you have to compare y-5 with y+5.
The good years in the early 90s were due to the recession; reduced annual mileages and probably
reduced "high living".
>Taking the 10 year improvements,
>1960-1970 through to 1976-1986, accident rates fell by about 40% per decade.
>1977-1987 through to 1981-1991 accident rates fell by about 45% per decade although this isn't as
>consistent a reduction, even when averaged over 10 years.
>1982-1992 thought to 1986-1996 accident rates fell by about 50% per decade
Mainly due to the recession 1990 - 1993.
>1987-1997 through to 1989-1999 accident rates fell by about 45% per decade.
Same again.
>Beyond 1999 the method of stats collection changed but the figures are sitting somewhere abound
>35-45% for the two years I have. Yes the decade 1991 to 2001 is the worst in the period I have
>considered but this is only one period, there were worse periods in the 1960-1970 through
>1976-1986.
>Finally, there were four exceptional years in succession, 1991 - 12% fall, 1992 8% fall, 1993, 10%
>fall, 1994 7% fall. These four years occured after the magic 1991 "speedcamera" year. I would
>expect speed cameras to be placed first at the worst accident blackspots and then only later get
>installed at the less "accident prone" sites so although 1991 may only have had a few speed
>cameras, they were probably sited at the places which could give the best improvement. I remember
>the Bham New Road that regularly used to have serious accidents because of cars jumping red lights.
>You could almost overnight see the difference once the traffic light cameras got installed. There
>were a few speed cameras on this road but, in general, if people drove at speeds where they could
>stop for a red light, then then generally wouldn't be exceeding the speed limit much anyway. (I
>rarely drive this road now and am not local so I don't know whether the accident improvements have
>been maintained but I hope so)
Speed cameras were trialed in 1992 and started to be used seriously in 1993.
--
Paul Smith Scotland, UK http://www.safespeed.org.uk please remove "XYZ" to reply by email speed
cameras cost lives