red lights

Peter Clinch said:

Quoted message said:

What methods have you used to observe them? I think a
good one would be to pick random individual cyclists and
follow them.

Methodolgy is very poor, because you'll have a sample base
of /a/ cyclist per follow.

I'm not testing the routes, I'm testing the people. I don't
need more than one sample per follow. Why is that poor if
enough follows are done over a long enough period of time? I
can see that the changing seasons and any changes to the
infrastructure might possibly have an effect to the type of
people cycling and on their behaviour ("summer madness"?)
but I would have thought those errors are no more serious
than any others you'd get with any other practicable
research. At the very least it's attempting to answer the
right question.

I'm not that bothered about it being small and slow if it
gets consistent results because it's interesting.

Quoted message said:

Much better to find some fantastically busy lights and
observe from a distance (so you don't affect the
behaviour) and count a goodly sample base going through
over varied times of day.

Fundamentally flawed, not answering the question we're
asking. Cyclists that jump lights tend not to jump every
time they get the opportunity so there would be a massive
error on a "busy" junction because it's the quiet places
that get jumped the most, and it would be too difficult and
subjective to judge what counts as a typical jumper junction
and a typical opportunity to jump the light.

Ideally, I would like to know what proportion of London
cyclists NEVER go through a red light. "Never" is
impossible to know so I'll settle for no jumps within a
limited distance.

~PB

Gawnsoft said:

Quoted message said:

Nobody in their right mind believes that label because
they know it doesn't make sense. They don't need to go
by the Daily Mail's label. They see the cyclists for
themselves out in the real world. They see them break
the law but they see that most of these law-breakers
are perfectly respectable-looking men and women causing
no harm.

So you can testify to the inner thoughts of strangers can
you? Including the ones who end up writing in to the
papers, or who write newspaper coumns, about 'louts'?

That is easy when you read and hear the inconsistencies and
hypocrisies the individuals reveal over time.

~PB

What relevance does a measure of the percentage of
cyclists who NEVER go through a red light have? The same
as the percentage of drivers who NEVER exceed any speed
limit? Surely it is the behaviour of populations that is
of most interest?

I had occasion to cycle around Hull yesterday evening for 3
hours and saw NO cyclists ride across a junction against a
red light, although plenty went past the stop line to avoid
being mown down in the traffic light GP. I encountered over
30 other cyclists in this time. Two cyclists left the road
and crossed via a signalised pedestrian crossing that was
showing green. In comparison at almost every set of signals
that I came across which were changing at least one car went
through against the red signal, so potentially coming into
conflict with all those setting of on amber on the traffic
light GP. I would guess many more would have if the
occasional driver had not actually stopped so preventing
those behind from running the red light. I.e. when measuring
behaviour we need to recognise that cyclists have much more
opportunity to run lights then cars stuck in traffic queues.

There are plenty of signals in Hull that fail to detect
cyclists so leaving one little option but to go though on
red when the junction is empty, get off an walk (assuming on
can actually get of the highway safely) or wait 5 minutes or
more until a car or buss comes along. Again cars don't have
this problem and I would bet if they did drivers would
similarly go though against a red signal.

The basic problem is that you seem to be concerned that
different sets of 'rules' are adopted by different sets of
road users. The basic 'rule' is that it is OK for a driver
to run a red light if this has just changed, but not for a
cyclist to ride through an empty junction, even if the
lights don't actually change in response to a waiting
cyclist. Also, how about taking a wider perspective and also
looking at the percentage of pedestrians who fail to act
according to traffic signals and attempt to identify the
'rules' that govern how they behave?

The bottom line is that if a cyclist wishes to place their
own safety at risk, that I up to them. However, no one has
the right to put the safety of others at risk and given the
danger posed by motor vehicles it is absolutely necessary
for drivers to drive to a high standard at all times. This
is why driving a motor vehicle is a privilege, controlled
through licence, not a right as is walking, riding a horse
or cycling on a public road. Unfortunately, so many have
been granted the privilege of driving on the public road
that this impacts on the utility value of the roads system
to those who supposedly use it by right. For example, the
inconvenience caused by all those traffic lights erected in
order to control the flow of motor vehicles...

"Pete Biggs" <[email hidden]> wrote in message news:<[email hidden]>...

Quoted message said:

I'm not testing the routes, I'm testing the people. I
don't need more than one sample per follow. Why is that
poor if enough follows are done over a long enough period
of time?

It's not poor. It will be statistically significant if you
follow enough people. Where bias might creep in is in
selecting the subjects to follow. You need to have an
objective repeatable method.

In London I suspect you'll find well over 50% regularly go
through red lights.

--
Dave...

Pete Biggs said:

I'm not testing the routes, I'm testing the people. I
don't need more than one sample per follow. Why is that
poor if enough follows are done over a long enough period
of time?

Because your sample base will be too small to create
meaningful data. You'll test a few people very thoroughly,
but the sample base may mean you quite by chance latch onto
a subset of cyclists with abnormal behaviour. You need lots
of people for the data to be credible, that's a simple fact
of statistics.

Quoted message said:

I'm not that bothered about it being small and slow if it
gets consistent results because it's interesting.

But if it's small it is rather by definition not going to be
useful data you can conclude anything from. c.f. "anecdotal
evidence" in the You Know What threads.

Quoted message said:

Fundamentally flawed, not answering the question we're
asking. Cyclists that jump lights tend not to jump every
time they get the opportunity so there would be a massive
error on a "busy" junction because it's the quiet places
that get jumped the most, and it would be too difficult
and subjective to judge what counts as a typical jumper
junction and a typical opportunity to jump the light.

Okay, you need a representative set of different junctions,
and a large sample base of cyclists. Without the former
you'll be asking the wrong question, as you suggest, but
without the latter you won't get any answers worth anything.
I suggested the above as it's possible to do as personal
interest research and get /some/ info back, as opposed to a
visit from the BiB for stalking people! ;-)

Pete.
--
Peter Clinch University of Dundee Tel 44 1382 660111 ext.
33637 Medical Physics, Ninewells Hospital Fax 44 1382 640177
Dundee DD1 9SY Scotland UK net [email hidden]
dundee.ac.uk~pjclinchOpen ↗

Howard said:

What relevance does a measure of the percentage of
cyclists who NEVER go through a red light have?

Revelant if zero-tolerance policing (for the offence) was
proposed so you'd know what proportion of cyclists would be
affected. Also interesting to contrast any difference
between this group and the average London cyclist.

Quoted message said:

The same as the percentage of drivers who NEVER exceed any
speed limit? Surely it is the behaviour of populations
that is of most interest?

I'm not saying what I want to find out is the MOST important
thing. It's just something that I think would be fascinating
to delve into further and it's a point which has been
disgussed on this newsgroup without much statistical
evidence presented.

Quoted message said:

I had occasion to cycle around Hull yesterday evening for
3 hours and saw NO cyclists ride across a junction against
a red light

Try that in London. You'll most likely see a cyclist jump a
light within 3 minutes if you beamed down to London in the
rush hour -- which is where I've been specifically referring
to all along.

~PB

Peter Clinch said:

Because your sample base will be too small to create
meaningful data. You'll test a few people very thoroughly,
but the sample base may mean you quite by chance latch
onto a subset of cyclists with abnormal behaviour. You
need lots of people for the data to be credible, that's a
simple fact of statistics.

The sample base would be built up over a period of weeks to
months. I think it would be large enough that the chances of
it being swayed by abnormal behaviour would be slim. I don't
pretend the research will be scientific enough to be taken
deeply seriously on its own, it would just be a starter. I
don't accept that it would be totally useless and
meaningless, however.

Quoted message said:

Quoted message said:

I'm not that bothered about it being small and slow if it
gets consistent results because it's interesting.

But if it's small it is rather by definition not going to
be useful data you can conclude anything from. c.f.
"anecdotal evidence" in the You Know What threads.

Call it anecdotal evidence but it will be more convincing
than any anecdotal evidence I have at the moment,
especially if people can repeat it for themselves and get
consistent results. That may then inspire them to seek/do
better research.

/snip

Quoted message said:

Okay, you need a representative set of different
junctions, and a large sample base of cyclists.

Deciding what counts as a representative junction is
impossible until you at least do some initial research
first. The kind I'm suggesting might be useful for that.
Also, some clever adjustment would be required to take
account for the fact that offenders don't offend each time,
even at "typical" places.

Quoted message said:

I suggested the above as it's possible to do as personal
interest research and get /some/ info back, as opposed to
a visit from the BiB for stalking people! ;-)

What's the BiB? And can they do anything about those bikers
who try and get a free ride off my slipstream :-)

~PB

Pete Biggs said:

Deciding what counts as a representative junction is
impossible until you at least do some initial
research first.

By "representative" I mean typical of many sites. So
Hangar Lane Gyratory is not typical. A random pelican
crossing on Finchley High Road probably is, as would be
the crossroads/lights where Gravel Hill meets
Bexleyheath Broadway. Hyde Park Corner is out (you'll be
happy to know!).

Quoted message said:

useful for that. Also, some clever adjustment would be
required to take account for the fact that offenders don't
offend each time, even at "typical" places.

No clever adjustment needed. That's why you have (and indeed
need) a large sample base which itself implies a fair
stretch of time to collect it.

Quoted message said:

What's the BiB?

Boys in Blue.

Quoted message said:

And can they do anything about those bikers who try and
get a free ride off my slipstream :-)

You obviously need a recumbent with a tailfairing, and then
they'll probably have a lot of trouble doing so... ;-)

Pete.
--
Peter Clinch Medical Physics IT Officer Tel 44 1382 660111
ext. 33637 Univ. of Dundee, Ninewells Hospital Fax 44 1382
640177 Dundee DD1 9SY Scotland UK net
[email hidden] dundee.ac.uk~pjclinchOpen ↗

On Wed, 2 Jun 2004 04:03:31 +0100, "Pete Biggs"
<[email hidden]> wrote (more or less):

Quoted message said:

Gawnsoft said:

On Sun, 30 May 2004 17:27:14 +0100, "Pete Biggs"
<[email hidden]> wrote (more or less):

Quoted message said:

I don't believe you have a near miss every time you go
through the pinch point.

Never claimed I have had.

Sorry for the exaggeration! See, we all do it :-)

Seriously, I apologise for misinterpreting what you said to
use it for my generalisation. I honestly thought you were
claiming to have very frequent near misses.

I did. And I was specific about them. And specific about how
far from my home they happened.

I only claimed /one/ 'pinch point' near miss.

Quoted message said:

The fact remains that people do often exaggerate when
moaning about people from a different "group".

--
Cheers, Euan Gawnsoft: gawnsoft.co.srgawnsoft.co.srOpen ↗
Symbian/Epoc wiki: html.dnsalias.nethtml.dnsalias.netOpen ↗ Smalltalk
links (harvested from comp.lang.smalltalk)
html.dnsalias.netsmalltalkOpen ↗

"Peter Clinch" <[email hidden]> wrote in message
"]news:[email hidden]...

I said:

Quoted message said:

some clever adjustment would be required to take account
for the fact that offenders don't offend each time, even
at "typical" places.

I mean they don't offend each time they have an opportunity.

Quoted message said:

No clever adjustment needed. That's why you have (and
indeed need) a large sample base which itself implies a
fair stretch of time to collect

it.

Sorry, I still don't understand how that answers my
question. Please, I need more explicit help! Just having a
large sample base doesn't seem to solve all the problems
in my mind.

What exactly is being counted in your suggestion? The
proportion of cyclists visiting representative sites who
offend? Or the proportion of only those who have an actual
/opportunity/ to offend that do so? The former is not what I
want to know. The latter would mean difficult and subjective
judgments for the researcher.

For the former, yes the "opportunities" would even out if
the sample base was large, but you'd still just arrive at an
average figure for a *single* average representative site
that would include all the cases where there was no
opportunity to offend. That's wildly different from what I'm
asking. I want to know how many offend when they have a much
larger opportunity to do so. You would need to multiply your
figure by something. What?

(I take the point about sampling at more than one site,
but I assume you'd just average them instead of doing
anything incredibly sophisticated with the data?)

~PB

"Dave Kahn" <[email hidden]> wrote in message
"]news:[email hidden]...

Quoted message said:

It's not poor. It will be statistically significant if you
follow enough people. Where bias might creep in is in
selecting the subjects to follow. You need to have an
objective repeatable method.

Thanks Dave. I appreciate the difficulty with selecting
subjects and the whole thing being objective. This has can
only be an informal non-scientific thing, for me, then --
just some glorified anecdotal evidence gathering.

I'm thinking of simply following anyone who happens to be
the next nearest cyclists to me wherever I am, whoever they
are, following on with the next cyclists ASAP after I'm
finished with them. I'm fast enough to keep up with nearly
everyone who rides a bike in inner London, but those who are
too fast (or reckless) will be excluded by default.
Otherwise I'd stop after the distance limit was up or a red
light jumped. Also excluded are pavement cyclists, and maybe
pedicabs (? not sure: they're so incredibly slow ...and
different).

I've tried this out a bit already and it's a recipe for the
most random bike rides I've ever done.

~PB

Quoted message said:

What exactly is being counted in your suggestion? The
proportion of cyclists visiting representative sites who
offend? Or the proportion of only those who have an actual
/opportunity/ to offend that do so? The former is not what
I want to know. The latter would mean difficult and
subjective judgments for the researcher.

What about setting up shop on a T junction, and counting
those that travel left to right aross the bar. Jumping a
light like this should be easily possible and one of the
safest lights to jump, so we can assume that a 'jumper' will
always have the opportunity and will reveal himself here.
Anyone waiting is either not a jumper, or knows that the
light is going green very soon. Given the relative safety in
which this type of junction can be crossed at red, it is IMO
likely that jumpers will go for it however small the delay.

This has the advantage of bathering data more quickly, is
vaguely valid - at least for that light :-) - but
unfortunately won't give you a decent bike ride.

Pete Biggs said:

Sorry, I still don't understand how that answers my
question. Please, I need more explicit help! Just having a
large sample base doesn't seem to solve all the problems
in my mind.

It does solve a lot of statistical problems. For example, if
people don't offend all the time and you miss them because
they've Been Good, you'll maybe catch them tomorrow when
they're not. Or later on, if they're in a hurry to get to
work and relaxed coming home (or vice versa!). If they never
stop at the junction you're parked at, that's where your
next couple of "representative" junctions may well catch
them over time.

Large sample bases give you a reasonable view of how a whole
population will behave, including unusual examples but
concentrating on the common ones (because there are more of
them, because they're common).

Following people takes a lot more time so your sample is
smaller (so less useful) and you also run the risk that your
observation modifies their behaviour because you're on a
nearby vehicle. What if they see other cyclists as a race,
for example? Hence the suggestion to observe from where you
can't really be seen to be observing (say, a cafe on a
corner with lights).

Quoted message said:

What exactly is being counted in your suggestion?

That's up to your experiment design. Most flexible is you
count everything you can think of and then you can work the
data at your leisure over time.

The primary useful bit in the above is that the cyclists
come to you in the course of their natural riding and you
see more of them. You'll build up a dataset quicker, and
because you see more of them you'll observe a greater range
of behaviour. Riding behind someone you see far less cycle
traffic, because half of it'll be going in the same
direction and if you're actively looking after your own
safety you'll have plenty of opportunity to lose the real
fools and miss folk going the other way.

Pete.
--
Peter Clinch Medical Physics IT Officer Tel 44 1382 660111
ext. 33637 Univ. of Dundee, Ninewells Hospital Fax 44 1382
640177 Dundee DD1 9SY Scotland UK net
[email hidden] dundee.ac.uk~pjclinchOpen ↗

Peter Clinch said:

It does solve a lot of statistical problems. For example,
if people don't offend all the time and you miss them
because they've Been Good, you'll maybe catch them tomorrow
when they're not. Or later on, if they're in a hurry to get
to work and relaxed coming home (or vice versa!). If they
never stop at the junction you're parked at, that's where
your next couple of "representative" junctions may well
catch them over time.

Ideally you want a mix of methodologies to fully understand
what's going on. Statistics from a single crossing.
Statistics from individual cyclists. Different times of
day, year, weather. In-depth interviews with a cross
section of people.

Erm - what was the originial research question exactly?

Though I'm normally not a red light jumper I sometimes
make an exception for the pedestrian crossing on Blackness
Road, if I can see the ped who pushed the button has
already crossed before it turned red and there isn't
another ped in sight.

Roos

Peter Clinch said:

Pete Biggs said:

Sorry, I still don't understand how that answers my
question. Please, I need more explicit help! Just having
a large sample base doesn't seem to solve all the
problems in my mind.

It does solve a lot of statistical problems. For example,
if people don't offend all the time and you miss them
because they've Been Good, you'll maybe catch them
tomorrow when they're not. Or later on, if they're in a
hurry to get to work and relaxed coming home (or vice
versa!). If they never stop at the junction you're parked
at, that's where your next couple of "representative"
junctions may well catch them over time.

That's only one of the minor problems. The major one is
judging what counts as a fair opportunity to offend. The
light has to be red at the very least. Then you've got to
decide if they were able to get to the stop line through any
queing traffic, etc, etc, etc. Despite all the flaws, I
think simply giving them the opportunity to do whatever they
want within a long limit (say 2 miles*) solves that problem
to a large extent.

* That's a lot in inner London, would have to be longer
elsewhere.

Quoted message said:

Large sample bases give you a reasonable view of how a
whole population will behave, including unusual examples
but concentrating on the common ones (because there are
more of them, because they're common).

Yes I know that but it doesn't matter how large the sample
base is unless you're sampling the necessary information.

Quoted message said:

Following people takes a lot more time so your sample is
smaller (so less useful) and you also run the risk that
your observation modifies their behaviour because you're
on a nearby vehicle. What if they see other cyclists as a
race, for example? Hence the suggestion to observe from
where you can't really be seen to be observing (say, a
cafe on a corner with lights).

It is a risk but only a small one in practice because they
tend not to /see/ the the following cycist if he keeps his
distance (I have tested the method) let alone be influenced
by him. Cyclists in London rarely look directly behind them.
When they do look behind, it tends to be a quick glance into
the near distance, to an angle.

Quoted message said:

Quoted message said:

What exactly is being counted in your suggestion?

That's up to your experiment design. Most flexible is you
count everything you can think of and then you can work
the data at your leisure over time.

To answer my question, the necessary recording and
computation would be extremely complex with this method.

Quoted message said:

The primary useful bit in the above is that the cyclists
come to you in the course of their natural riding and you
see more of them. You'll build up a dataset quicker, and
because you see more of them you'll observe a greater
range of behaviour.

It's no good getting loads of samples if you can't sort out
all the other problems.

Quoted message said:

Riding behind someone you see far less cycle traffic,
because half of it'll be going in the same direction

I only need to see one cycle at a time.

Quoted message said:

and if you're actively looking after your own safety
you'll have plenty of opportunity to lose the real fools
and miss folk going the other way.

The real fools jump lights pretty quickly so fortunately the
test is over quickly. I don't tend to loose the berks where
there are no lights as I catch them up after I've negotiated
the danger, etc.

~PB

Roos Eisma said:

Erm - what was the originial research question exactly?

"Can you demonstrate that with enough data you can prove
anything you set out to through statistical methods?" ;-/

Pete.
--
Peter Clinch Medical Physics IT Officer Tel 44 1382 660111
ext. 33637 Univ. of Dundee, Ninewells Hospital Fax 44 1382
640177 Dundee DD1 9SY Scotland UK net
[email hidden] dundee.ac.uk~pjclinchOpen ↗

"Peter Clinch" <[email hidden]> wrote in message
"]news:[email hidden]...

Quoted message said:

Roos Eisma said:

Erm - what was the originial research question exactly?

"Can you demonstrate that with enough data you can prove
anything you set out to through statistical methods?" ;-/

No you can't. But with LITTLE enough data you can deduce
anything :-)
--
Mark South: World Citizen, Net Denizen

Peter Clinch said:

Quoted message said:

I'm not testing the routes, I'm testing the people. I
don't need more than one sample per follow. Why is that
poor if enough follows are done over a long enough period
of time?

Because your sample base will be too small to create
meaningful data. You'll test a few people very thoroughly,
but the sample base may mean you quite by chance latch onto
a subset of cyclists with abnormal behaviour. You need lots
of people for the data to be credible, that's a simple fact
of statistics.

Hi Peter

But it's the individual cyclist who jusps red lights etc who
cives cyclists in general a bad name.

James

Roos Eisma said:

Though I'm normally not a red light jumper I sometimes
make an exception for the pedestrian crossing on Blackness
Road, if I can see the ped who pushed the button has
already crossed before it turned red and there isn't
another ped in sight.

Hi Roos

Even more annoying is the pavement cyclist who presses the
button then rides across to the opposite pavement and
continues on his merry way. Last bank holiday Monday I, and
at least 20 cars, were held up by a young teenager crossing
the road via the crossing. I might had some sympathy with
the guy had he actually got off his bike.

James

James Hodson said:

But it's the individual cyclist who jusps red lights etc
who cives cyclists in general a bad name.

Well, yes, but I think it's Pete's idea to find out how
individual the individuals are, IYSWIM

Pete.
--
Peter Clinch Medical Physics IT Officer Tel 44 1382 660111
ext. 33637 Univ. of Dundee, Ninewells Hospital Fax 44 1382
640177 Dundee DD1 9SY Scotland UK net
[email hidden] dundee.ac.uk~pjclinchOpen ↗