Safe Speed Forums

Right. I've had enough and I'm going into battle.

It is OUTRAGEOUS the the new DfT report into speed camera effectiveness does not cater for the "regression to the mean" error:
http://www.safespeed.org.uk/rttm.html

I have written an open letter to Professor Heydecker:
http://www.safespeed.org.uk/heydecker2.html

And I have sent it out with a PR to over 700 motoring related journalists and editors:
http://www.safespeed.org.uk/pr126.html

The blue touchpaper has been lit, and I'm waiting for the fireworks. Funny thing is, most of this went out yesterday and it's been a pretty quiet day.

_________________
Paul Smith
Our scrap speed cameras petition got over 28,000 sigs
The Safe Speed campaign demands a return to intelligent road safety

Numberwatch joins in!

http://www.numberwatch.co.uk/2004_June.htm#speed

_________________
Paul Smith
Our scrap speed cameras petition got over 28,000 sigs
The Safe Speed campaign demands a return to intelligent road safety

Of course they didnt sift through the 100 sites and choose the 10 that would give them a 35% reduction did they.
What if they worked out the reduction for all of the 100 sites?
What if they did it for 10,000 sites?
What if they chose the 10 sites out of the 100 that gave the least reduction?
What if they did this over a period of 10 years?
etc.......................

Looks like a very suspicious set of figures to me and doesnt prove against the report.

_________________
It's Champion Man

Eh

Their table is generated from a set of random numbers, it's designed to demonstrate the regression to the mean phenomenon.

Isn't it

Exactly that. Of course JB has chosen the perfect method to demonstrate the effect with the right sort of numbers with the right sort of distribution.

_________________
Paul Smith
Our scrap speed cameras petition got over 28,000 sigs
The Safe Speed campaign demands a return to intelligent road safety

Oh, I see, a set of random numbers. Made them up in other words.

_________________
It's Champion Man

Don't be a nit wit - most accidents aren't predictable either.

If you want "real world data", try this from Professor John Adams: "There was a famous study done in Sweden of over 2,500 unsignalled junctions and they discovered that the accident black spots without any treatment at all in the ?after? period had decreased by over 50% Now that's the sort of order of magnitude of increased safety that is often claimed for black spot treatments."

_________________
Paul Smith
Our scrap speed cameras petition got over 28,000 sigs
The Safe Speed campaign demands a return to intelligent road safety

Yes indeed "made up", but try it for yourself.

1. Write the numbers 1 to 10 on 10 scraps of paper.
2. Put them in a hat and pick them out in random order. Write them on a sheet of paper as a column of numbers in this order. At the head of this column write the title "Before".
3. Now jumble the numbers up and pick them out again. Write these in as column two in the new random order, and title this column "After".

Ok, now you have a small simulation of a system where the same numbers are moving randomly from one location to another. Think of these as KSIs at 10 sites.

Now lets pick the worst (say) 3 of them, ie the sites where 10,9 and 8 land in the first column. Now look at the second column and see how much the KSIs have reduced by. The odds are that you will now have seen a substantial decrease in accidents at the "treated" sites, yet in reality the total number has stayed exactly the same.

I've just tried it and my KSI's dropped from 27 to 12 - a 56% reduction!

And that is regression to the mean. No smoke, no mirrors, no trap doors, no hidden strings. What answer do you get?

_________________
CSCP Latin for beginners...
Ticketo ergo sum : I scam therefore I am!

This just gone to 710 motoring connected journalists, editors and politicians:

PR127: Professors agree speed camera report is fraudulent

NEWS: For immediate release

Speaking on Radio 5 live on 15th June 2004, Paul Smith of the Safe
Speed road safety campaign, revealed that new government figures
claiming success for speed cameras included a serious statistical
error. Paul followed this up with a press release and an open letter
to the lead author of the offending report, Professor Heydecker.

Paul Smith, founder of the Safe Speed road safety campaign said:
"Speed cameras in the UK are an unmitigated disaster. Now we seem to
be suffering, not only from the cameras themselves, but also from
layers of deceit as those who have faith in the camera system try to
paper over the spreading cracks with scientifically invalid
"evidence"."

Following Safe Speed's revelations many senior academics have joined
the complaints. They include:

Professor emeritus Mervyn Stone (UCL)
Professor emeritus John Brignell (Southampton)
Professor John Adams (UCL)
Professor Garel Rhys (Cardiff)
Professor Michael Silver (Cardiff)

Now "Numberwatch", a web site edited by Professor John Brignell, says:
"In particular, one Benjamin Heydecker, who rejoices in the title of
Professor in the Centre for Transport Studies University College
London, persistently renews this deception, although its fraudulent
nature has been pointed out to him over and over again. That is the
way the professorate behaves in the post-scientific age."

In a written report for the Radio 4 Today programme, Professor Mervyn
Stone said: "The three-year DfT report was released from a
politically-dictated embargo on June 15th. As I implied in my speed
camera judgement, its analysis of the data from the 24 police force
areas makes no quantitative allowance for regression to the mean.
Section G3 of the report gives reasons for thinking that the
established statistical phenomenon of regression to the mean will not
apply in full measure". I interpret this to mean that the authors
acknowledge that the estimates of savings of casualties and accidents
in the Executive Summary should be taken as upper bound estimates.
Will the public be made sufficiently aware of this qualification? Was
the minister?"

<ends>


Notes for editors.
====================

Comments regarding the DfT report on the Numberwatch web site:
http://www.numberwatch.co.uk/2004_June.htm#speed

Professor John Brignell's CV:
http://www.ecs.soton.ac.uk/~jeb/cv.htm

Numberwatch home page:
http://www.numberwatch.co.uk

Previous Safe Speed PR:
http://www.safespeed.org.uk/pr126.html

Safe Speed letter to Professor Heydecker:
http://www.safespeed.org.uk/heydecker2.html

Regression to the mean primer:
http://www.safespeed.org.uk/rttm.html

DfT Report: "The national safety camera programme: Three-year
evaluation report":
http://www.dft.gov.uk/stellent/groups/d ... 029194.pdf

DfT Road Safety Good Practice Guide:
http://www.dft.gov.uk/stellent/groups/d ... 64-06.hcsp

Professor John Adams (UCL) quoted in The Times, 25th June 2004
http://www.timesonline.co.uk/printFrien ... 90,00.html

Professors Garel Rhys (Cardiff) and Michael Silver (Cardiff) quoted
in MCN 23rd June 2004:

Professor Mervyn Stone (UCL) See...

http://www.bbc.co.uk/radio4/today/repor ... 0624.shtml

... Where the BBC are due to publish a report by Professor Mervyn
Stone including the quote above and much more information about speed
cameras and regression to the mean. Publication has been delayed.
Professor Stone's 11,000 word report is highly critical of the UK
speed camera programme.

_________________
Paul Smith
Our scrap speed cameras petition got over 28,000 sigs
The Safe Speed campaign demands a return to intelligent road safety

I followed the link above, and it appears that not only does Professor Stone believe that speed cameras save lives, he reckons the "four-coffin" rule should be scrapped and/or cameras made much less visible. So his report is not exactly a ringing endorsement of your fanatical anti-camera position.

That's effectively just spin. Wait 'til you read the report. Mervyn Stone says he'll publish it via UCL if the BBC drag their heels for too long.

In the meantime you can read all the other documents including my reply from:

http://www.safespeed.org.uk/stone.html

_________________
Paul Smith
Our scrap speed cameras petition got over 28,000 sigs
The Safe Speed campaign demands a return to intelligent road safety

Yes, I heard him say things on the wireless that were clearly pro-camera, and yet the spin here is that Stone is on-message with the SafeSpeed anti-camera position. There may be some truth in this 'regression to the mean' idea, which would be eliminated by longer sampling trends, but there is no way that Stone is a SafeSpeed ally in this discussion.

_________________
I stole this .sig

If you are dealing with true random numbers, any trend is bogus. Indeed, some ‘unlucky’ sites suffer a random run of accidents that could be misinterpreted over a short period as a trend, triggering installation of signage or a camera. Over a longer period with more samples, the trend for these sites would have been shown to be bogus, but this opportunity is removed because the ‘improvement’ is assumed to be due to the changes.

On the other hand, some sites suffer a run of accidents that are causally linked, usually due to some feature of the nearby road system. This is backed up by common sense – we all know of bad tee-junctions and poor cambers, blind bends and bad crossings etc. Such a site would show up over the longer term as a clear trend, not subject to ‘regression to the mean’ effect.

This would suggest that the threshold trigger for a camera installation should involve better statistics, but does not undermine the general case for having cameras. A massive amount of signage has been put up around the country to try to ameliorate the problem of ‘accident blackspots’, without much effect. Much of this signage has been intended to warn traffic and slow it down around the danger spot. Now we have signs that bite (or nip, at least) – Speed Cameras – and anybody who drives in north Wales or Peterborough can judge how well they work.

_________________
I stole this .sig

You're right in as much as Stone is not a Safe Speed ally. But he didn't cause any of my arguments to unravel and he did unravel the government's arguments. He also said nice things about the quality and scope of my work.

You need to read the report before you make any judgement. The BBC are still dragging their heels, but Professor Stone has agreed to publish the report via UCL as soon as possible.

_________________
Paul Smith
Our scrap speed cameras petition got over 28,000 sigs
The Safe Speed campaign demands a return to intelligent road safety

The Stone report is now available via UCL.

http://www.ucl.ac.uk/Stats/research/Resrprts/speed.pdf

The following Safe Speed Page has links to the report and other materials:

http://www.safespeed.org.uk/stone.html (includes Safe Speed reply)

The Beeb still have not published the Stone report, nor mine nor Robert's evidence documents. Don't know why.

_________________
Paul Smith
Our scrap speed cameras petition got over 28,000 sigs
The Safe Speed campaign demands a return to intelligent road safety