P
Paul Smith
Guest
On Wed, 2 Apr 2003 23:24:41 +0100, John Openshaw <[email protected]> wrote:
>>>>Most importantly, I suggest that "Within the range of influence of any imaginable degree of
>>>>increased speed enforcement (using current technologies and enforcement systems) the correlation
>>>>between free travelling speeds and impact speeds is between zero and negative, except perhaps at
>>>>some very special accident black spot sites."
>I'm not one for pots and pans either so I had a look at the web-site, in particular the
>international graphs. You note that the deaths are on an exponential scale and the numbers are
>decreasing. As the numbers decrease so it takes a smaller number of deaths to change things
>dramatically (at least to the eye), and its not easy or even meaningful to draw conclusions.
But the size of the system is big and increasing, (greater number of vehicle kms) and that leads to
a remarkable degree of consistency from year to year. For example the last four years (-> 2001) GB
fatality figures are:
3421, 3423, 3409, 3450 (i.e. within about 0.6% of average)
But for an earlier four years in 1980 -> 1983
6010, 5846, 5934, 5445
I'm convinced we're at the "bottom" of a trend reversal.
>For example, if you plot a hypothetical line (log or lin domain makes no difference), using any
>recognised line fitting technique, what is the confidence level at which your observations take
>place? In fact, I'm surprised to see that your in-depth numerical analysis omits error bars, as
>these could quite easily invalidate/validate everything you say.
Sadly, as of now, I don't know how to do that.
>As far as the web-site goes, you say, and I paraphrase: 'the graphs go wonky at a certain date, and
>this date coincides with speed cameras, ergo there is a link'. Unfortunately you do not prove
>statistically the graphs go wonky, and there is no evidence on your site of a causal link between
>speed cameras and fatal accidents.
I certainly don't say "ergo there is a link". I do say it's the only explanation I've got which fits
the observed effect, and I did predict the effect long before I found anything in the data to
support it.
I look forward to proving that the "graphs go wonky".
As for proving a causal link, that would be hard indeed even with unlimited resources, given the
proposed mechanism:
http://www.safespeed.org.uk/tiger.html
I certainly don't expect to be able to prove causality.
>So one question I would like answered as I cant see it is: You say that the death rates have either
>increased or decreased (according to speed kills campaigns) - what is the confidence interval for
>your hypothesis for all the international graphs? Is it 99%, 95% or what?
I'm working on all of this, John. I've purchased 4 text books about statistics, and I'm learning
stuff that I've forgotten from 30 years ago when I last used it. It certainly doesn't help that
Excel is a seriously **** statistics tool and tends to lead one astray. I just love Microsoft.
--
Paul Smith Scotland, UK http://www.safespeed.org.uk please remove "XYZ" to reply by email speed
cameras cost lives
>>>>Most importantly, I suggest that "Within the range of influence of any imaginable degree of
>>>>increased speed enforcement (using current technologies and enforcement systems) the correlation
>>>>between free travelling speeds and impact speeds is between zero and negative, except perhaps at
>>>>some very special accident black spot sites."
>I'm not one for pots and pans either so I had a look at the web-site, in particular the
>international graphs. You note that the deaths are on an exponential scale and the numbers are
>decreasing. As the numbers decrease so it takes a smaller number of deaths to change things
>dramatically (at least to the eye), and its not easy or even meaningful to draw conclusions.
But the size of the system is big and increasing, (greater number of vehicle kms) and that leads to
a remarkable degree of consistency from year to year. For example the last four years (-> 2001) GB
fatality figures are:
3421, 3423, 3409, 3450 (i.e. within about 0.6% of average)
But for an earlier four years in 1980 -> 1983
6010, 5846, 5934, 5445
I'm convinced we're at the "bottom" of a trend reversal.
>For example, if you plot a hypothetical line (log or lin domain makes no difference), using any
>recognised line fitting technique, what is the confidence level at which your observations take
>place? In fact, I'm surprised to see that your in-depth numerical analysis omits error bars, as
>these could quite easily invalidate/validate everything you say.
Sadly, as of now, I don't know how to do that.
>As far as the web-site goes, you say, and I paraphrase: 'the graphs go wonky at a certain date, and
>this date coincides with speed cameras, ergo there is a link'. Unfortunately you do not prove
>statistically the graphs go wonky, and there is no evidence on your site of a causal link between
>speed cameras and fatal accidents.
I certainly don't say "ergo there is a link". I do say it's the only explanation I've got which fits
the observed effect, and I did predict the effect long before I found anything in the data to
support it.
I look forward to proving that the "graphs go wonky".
As for proving a causal link, that would be hard indeed even with unlimited resources, given the
proposed mechanism:
http://www.safespeed.org.uk/tiger.html
I certainly don't expect to be able to prove causality.
>So one question I would like answered as I cant see it is: You say that the death rates have either
>increased or decreased (according to speed kills campaigns) - what is the confidence interval for
>your hypothesis for all the international graphs? Is it 99%, 95% or what?
I'm working on all of this, John. I've purchased 4 text books about statistics, and I'm learning
stuff that I've forgotten from 30 years ago when I last used it. It certainly doesn't help that
Excel is a seriously **** statistics tool and tends to lead one astray. I just love Microsoft.
--
Paul Smith Scotland, UK http://www.safespeed.org.uk please remove "XYZ" to reply by email speed
cameras cost lives