Passing thoughts

Buck said:

I have been considering the results of the recently posted brief on Ian
Walker's study of motorists passing cyclists
(http://www.drianwalker.com/overtaking/). I have come to realize that
there is some legitimacy to his work. Although, I think he missed an
important point.

Now that I've read not only other people posting on this, but the press
release and the managerial version of the publication (the full
publication is not yet out yet), I think this is junk science.

It's amusing to discuss, and the findings might even be right, but
there's no validity to the study itself.

I've outlined this in more detail here:
<http://journals.aol.com/mikekr/ZbicyclistsZlog/entries/2006/09/16/want-to-get-hit-by-a-car-wear-a-helmet/798>

Quoted message said:

Buck said:

I have been considering the results of the recently posted brief on Ian
Walker's study of motorists passing cyclists
(http://www.drianwalker.com/overtaking/). I have come to realize that
there is some legitimacy to his work. Although, I think he missed an
important point.

Now that I've read not only other people posting on this, but the press
release and the managerial version of the publication (the full
publication is not yet out yet), I think this is junk science.

Perhaps you will apply to be on the editorial board.

Quoted message said:

(http://www.drianwalker.com/overtaking/). I have come to realize that
there is some legitimacy to his work. Although, I think he missed an
important point.

There may be legitimacy to the concept, but there is *no* legitimacy to the
manner in which he performed his "study."

As I pointed out in an earlier thread-

"A sample size of ONE rider. That's right, ONE rider, who emulated
male & female by using a wig or not. This is not good science. It's not even
a good story.

It makes the case of over 1000 cars, which is fine, reasonable sample size.
But it draws conclusions based upon an unobserved single participant (who
just happens to be the author of the study, and who just happens to have
been hit TWICE during the study after having been passed by 2500 motorists.

Think about what that's saying. Let's assume he wore the helmet half the
time, and both times he was hit he was wearing the helmet. If that's the
case, then his rate of being physically hit by a passing car is ONCE EVERY
625 CARS!!! If this were typical, most of us wouldn't live long enough to
figure out the next version of Windows, Linux, or Apple OS-X.

Complete & total nonsense. It could be that the premise actually is true...
that cars may in fact actually pass closer to a helmeted cyclist than a
non-helmeted cyclist... but there is nothing in that experiment to support
that, due to the absurd flaws in how it was run. Other than to suggest that
the author is a terribly unskilled cyclist, that is. That conclusion is
inescapable."

Quoted message said:

When I move further left in my lane, passing behavior changes. Yes, the
relative position between me and the passing vehicle may be closer.
However, I find that people are more reluctant to pass me illegally.
The two most common illegal passes I encounter are passing in a
no-passing zone, and not moving completely into the other lane when
passing (lane sharing). It is the latter behaviour that my lane
position seems to affect most.

As you act more like a car, the cars will treat you more as they would treat
a car. Which is generally a good thing. I'm not a John Forrester type; I
don't believe that all cyclists are best-served by acting like a car. But it
is an advantage for the stronger riders, who can more easily take the lane
when required, with a minimal effect on the flow of traffic.

Quoted message said:

Pulling my kids' trailer around also seems to inspire the same results.
An added benefit of the trailer is the speed at which people pass.
Whether they are slowing to observe this unusual contraption on the
road or they have genuine concern about what may be inside, I don't
know. But I have fewer vehicles zooming past when the trailer is on
board. The added benefit is a place to hang my clothes for work!

Agree 100%. People are often terrified of pulling their kids in a trailer,
believing they're going to get run over. My experience hauling my kids
around was the opposite; people gave me a very comfortable amount of space,
much more than when I rode without the trailer. Yes, it shows that people
drive differently depending upon what they see on the road in front of them,
but it doesn't add validity to the experiment, because it has nothing to do
with *you* riding differently, which I suspect was the case with the
cyclist/author in the experiment (when helmeted vs non-helmeted).

--Mike-- Chain Reaction Bicycles
www.ChainReactionBicycles.com

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

Quoted message said:

I have been considering the results of the recently posted brief on Ian
Walker's study of motorists passing cyclists
(http://www.drianwalker.com/overtaking/). I have come to realize that
there is some legitimacy to his work. Although, I think he missed an
important point.

When I move further left in my lane, passing behavior changes. Yes, the
relative position between me and the passing vehicle may be closer.
However, I find that people are more reluctant to pass me illegally.
The two most common illegal passes I encounter are passing in a
no-passing zone, and not moving completely into the other lane when
passing (lane sharing). It is the latter behaviour that my lane
position seems to affect most.

Pulling my kids' trailer around also seems to inspire the same results.
An added benefit of the trailer is the speed at which people pass.
Whether they are slowing to observe this unusual contraption on the
road or they have genuine concern about what may be inside, I don't
know. But I have fewer vehicles zooming past when the trailer is on
board. The added benefit is a place to hang my clothes for work!

I would like to see him revisit the video data he collected and make
the comparison in passing technique. I've emailed him with this
suggestion.

-Buck

Mike Jacoubowsky said:

Quoted message said:

(http://www.drianwalker.com/overtaking/). I have come to realize that
there is some legitimacy to his work. Although, I think he missed an
important point.

There may be legitimacy to the concept, but there is *no* legitimacy to the
manner in which he performed his "study."

As I pointed out in an earlier thread-

"A sample size of ONE rider. That's right, ONE rider, who emulated
male & female by using a wig or not. This is not good science. It's not even
a good story.

I disagree with you on this. When more than one person collects data,
additional variation is introduced into the data collection process.
Using a single person to collect data is a way to eliminate the
variation caused by slight changes in measurement technique.

This is easily observable in any high school chemistry class. Ask a
group of students to read the volume in a graduated cylinder. You will
get a variety of answers because they all read the meniscus slightly
differently. This is despite clear instructions on the proper
methodology for reading the meniscus.

You may argue that the example is for students in a learning
environment. But scientists in the medical field use the same
reasoning. When I was involved in medical research, we assigned
specific technicians to specific tasks throughout the life of each
study we conducted. This was to prevent any variation in techniques. It
is easier to deal with a consistent error than a random error.

If we were to toss out all research in which one technician or one
researcher were responsible for all data collection, we would toss out
a great deal of research.

-Buck

Quoted message said:

If we were to toss out all research in which one technician or one
researcher were responsible for all data collection, we would toss out
a great deal of research.

We're talking about two entirely different things here. I hope.

One of the most-basic requirements for such studies is being able to
duplicate the conditions and verify the results. Most technicians &
researchers would be careful to use tools & situations that would minimize
variability in results if someone else were to conduct the experiment. When
such variability cannot be helped, the solution is to have a cross-section
of people or whatever involved, so that the bias or technique of an
individual does not unduly shape the results of the study.

So I agree, we shouldn't be tossing out all research done by just one
person. But we have to look at the FACTS presented in this case, few as they
may be. And the FACTS are that this guy had two collisions with cars,
apparently during the 1250 times he was passed by a car while wearing a
helmet.

I would have been dead long ago if those were the odds I faced each time I
got out on a bike. As would most people in this newsgroup.

Rarely do I feel so strongly about an issue that it agitates me this much.
But bad science applied to cycling, whether it be something like this (which
affects how people view their safety on the road) or scaring people
literally off their seats by claiming they'll become impotent... yes, it
really bugs me that stories like those get media attention.

--Mike Jacoubowsky
Chain Reaction Bicycles
www.ChainReaction.com
Redwood City & Los Altos, CA USA

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

Quoted message said:

Mike Jacoubowsky said:

Quoted message said:

(http://www.drianwalker.com/overtaking/). I have come to realize that
there is some legitimacy to his work. Although, I think he missed an
important point.

There may be legitimacy to the concept, but there is *no* legitimacy to
the
manner in which he performed his "study."

As I pointed out in an earlier thread-

"A sample size of ONE rider. That's right, ONE rider, who emulated
male & female by using a wig or not. This is not good science. It's not
even
a good story.

I disagree with you on this. When more than one person collects data,
additional variation is introduced into the data collection process.
Using a single person to collect data is a way to eliminate the
variation caused by slight changes in measurement technique.

This is easily observable in any high school chemistry class. Ask a
group of students to read the volume in a graduated cylinder. You will
get a variety of answers because they all read the meniscus slightly
differently. This is despite clear instructions on the proper
methodology for reading the meniscus.

You may argue that the example is for students in a learning
environment. But scientists in the medical field use the same
reasoning. When I was involved in medical research, we assigned
specific technicians to specific tasks throughout the life of each
study we conducted. This was to prevent any variation in techniques. It
is easier to deal with a consistent error than a random error.

If we were to toss out all research in which one technician or one
researcher were responsible for all data collection, we would toss out
a great deal of research.

-Buck

Quoted message said:

Quoted message said:

"A sample size of ONE rider. That's right, ONE rider, who emulated
male & female by using a wig or not. This is not good science. It's not even
a good story.

I disagree with you on this. When more than one person collects data,
additional variation is introduced into the data collection process.
Using a single person to collect data is a way to eliminate the
variation caused by slight changes in measurement technique.

Wow, so we have really great data applicable to ONE person.

The problem isn't that ONE person collected the data - the problem is that
ONE person was the experiment from which the data was collected.

Chris Neary
[email hidden]

"Prize the doubt, low kinds exist without"
- Inscription at Ramsmeyer Hall, Ohio State University

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

Quoted message said:

Mike Jacoubowsky said:

"A sample size of ONE rider. That's right, ONE rider, who emulated
male & female by using a wig or not. This is not good science. It's not
even
a good story.

I disagree with you on this. When more than one person collects data,
additional variation is introduced into the data collection process.
Using a single person to collect data is a way to eliminate the
variation caused by slight changes in measurement technique.

This is easily observable in any high school chemistry class. Ask a
group of students to read the volume in a graduated cylinder. You will
get a variety of answers because they all read the meniscus slightly
differently. This is despite clear instructions on the proper
methodology for reading the meniscus.

Yes, there is variance added when more than one person collects data. The
proper statististical techniques can separate out this variance.

But the important thing is that this variance per person is very necessary
in a study like this. We don't know what cues the experimenter -- who knows
what he expects to find -- might have been giving off. There's no control on
this. You mentioned chemistry, so you are surely aware that medical studies
of new drugs are, ideally, done double-blind to avoid this type of problem.

One thing that we do know is that he got hit twice in 200 miles of riding
during this study. Does that sound typical of anybody?

This seems like a nice pilot study, just to shake down the methodology and
see if there might be something there. But it should have been followed up
by an adequately controlled study that might have actually told us
something.

Mike Kruger
--
"President Washington, President Lincoln, President Wilson, President
Roosevelt have all authorized electronic surveillance on a far broader
scale." --Attorney General Alberto Gonzalez

Buck said:

If we were to toss out all research in which one technician or one
researcher were responsible for all data collection, we would toss out
a great deal of research.

Research should be evaluated on its own merits. You'll find that it
is people with an agenda who will sieze upon a single element of some
study, and generalize it to all the studies with whose results they
disagree, and so "toss out" all they find opposing them.

If they truly suspected there was an effect from a single evaluator,
rather than strut out an opinion, they (and we) would be better served
if they repeated the experiment, changing only the number of
evaluators. This of course would be work, and blather is much the
simpler course.

Mike Jacoubowsky said:

One of the most-basic requirements for such studies is being able to
duplicate the conditions and verify the results. Most technicians &
researchers would be careful to use tools & situations that would minimize
variability in results if someone else were to conduct the experiment. When
such variability cannot be helped, the solution is to have a cross-section
of people or whatever involved, so that the bias or technique of an
individual does not unduly shape the results of the study.

One of the sticking points in my mind was the method used to verify
lane position. I contacted Dr. Walker and asked him a few questions
about his methodology. To control his lane position, he mounted a laser
pointer on his bike and set the angle to achieve the distances noted in
the already-published charts. While he did say that it was surprisingly
easy to maintain lane position with this method, I would not want to
follow the same methodology as it would take my focus away from the
traffic. I did not ask about his accidents.

Quoted message said:

So I agree, we shouldn't be tossing out all research done by just one
person. But we have to look at the FACTS presented in this case, few as they
may be. And the FACTS are that this guy had two collisions with cars,
apparently during the 1250 times he was passed by a car while wearing a
helmet.

I would have been dead long ago if those were the odds I faced each time I
got out on a bike. As would most people in this newsgroup.

At this point, I think that many of the questions brought about by the
brief will be answered when the full paper is publsihed. As to the
accidents, perhaps his methodology contributed to them, perhpas they
did not. I know that when bad things happen to me, they tend to happen
in quick succession.

Quoted message said:

Rarely do I feel so strongly about an issue that it agitates me this much.
But bad science applied to cycling, whether it be something like this (which
affects how people view their safety on the road) or scaring people
literally off their seats by claiming they'll become impotent... yes, it
really bugs me that stories like those get media attention.

What really bothers me about this story is that it was released to the
media BEFORE the full paper was published. Sensationalism rarely does
anyone any good. I will withhold my judgement on the quality of the
science until the full paper is published. However, I do believe the
reporters responsible for those stories should be forced to ride a few
hundred miles in downtown NY traffic without a helmet.

-Buck

Chris Neary said:

Wow, so we have really great data applicable to ONE person.

The problem isn't that ONE person collected the data - the problem is that
ONE person was the experiment from which the data was collected.

No. We have data regarding 2,300 vehicles passing a control. The
passing vehicles were the experiment, the cyclist was the control. The
control conditions were changed to see what the experiment participants
(although they didn't know they were participating) would do.

-Buck

Quoted message said:

Research should be evaluated on its own merits. You'll find that it
is people with an agenda who will sieze upon a single element of some
study, and generalize it to all the studies with whose results they
disagree, and so "toss out" all they find opposing them.

If they truly suspected there was an effect from a single evaluator,
rather than strut out an opinion, they (and we) would be better served
if they repeated the experiment, changing only the number of
evaluators. This of course would be work, and blather is much the
simpler course.

Absolutely agree 100%.

But... the study that's needed is a bit more broad-based than whether
wearing a helmet makes a difference or not. What's needed is a study that
simply asks (and answers) the question- "What are the behaviours while
riding a bike that influence your safety vs cars?"

--Mike Jacoubowsky
Chain Reaction Bicycles
www.ChainReaction.com
Redwood City & Los Altos, CA USA
<[email hidden]> wrote in message
news:[email hidden]...

Quoted message said:

Buck said:

If we were to toss out all research in which one technician or one
researcher were responsible for all data collection, we would toss out
a great deal of research.

Research should be evaluated on its own merits. You'll find that it
is people with an agenda who will sieze upon a single element of some
study, and generalize it to all the studies with whose results they
disagree, and so "toss out" all they find opposing them.

If they truly suspected there was an effect from a single evaluator,
rather than strut out an opinion, they (and we) would be better served
if they repeated the experiment, changing only the number of
evaluators. This of course would be work, and blather is much the
simpler course.

Mike Jacoubowsky said:

But... the study that's needed is a bit more broad-based than whether
wearing a helmet makes a difference or not. What's needed is a study that
simply asks (and answers) the question- "What are the behaviours while
riding a bike that influence your safety vs cars?"

I'm not sure what you have in mind there.

Yes, I think there is value in having Walker's passing distance study
repeated using more riders. But it's not surprising that he did it the
way he did, with just himself as a subject. Getting funding to study
such a thing is probably not easy; so rather than hire a small army of
college students to do the riding, he did it himself.

FWIW, I suspect there will be follow up studies trying to verify his
results. And I suspect a very well-funded study by the Harborview
Institute will claim to disprove it entirely, no matter what everyone
else finds! ;-)

But your more general question, "What are the behaviors while riding a
bike that influence your safety vs. cars?" - how exactly would you
propose to study this?

ISTM that you can't send out a gaggle of cyclists and ask them to all
ride differently. You'll get a multivariate mess of folks riding slow,
fast, close to the curb, away from the curb, straight and steady,
weaving, legally, illegally, etc. etc. If you can somehow measure
results, you'll have a hellacious statistical problem trying to
separate out the influences.

ISTM that it's more efficient to examine one variable at a time, which
is what Walker did. And it makes some sense to examine the factor that
is touted most loudly as a safety tool. I look forward to the
follow-up studies.

- Frank Krygowski

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

Quoted message said:

Chris Neary said:

Wow, so we have really great data applicable to ONE person.

The problem isn't that ONE person collected the data - the problem is
that
ONE person was the experiment from which the data was collected.

No. We have data regarding 2,300 vehicles passing a control. The
passing vehicles were the experiment, the cyclist was the control. The
control conditions were changed to see what the experiment participants
(although they didn't know they were participating) would do.

Your argument would be valid if the "control" was nonreactive -- e.g. a
lamp-post that would always be a lamp-post regardless of the hypothesis,
regardless of skill, and even with a wig would still be a lamp-post.

That doesn't apply in this case because the experimenter is potentially
reactive.

--
"President Washington, President Lincoln, President Wilson, President
Roosevelt have all authorized electronic surveillance on a far broader
scale." --Attorney General Alberto Gonzalez

Quoted message said:

Your argument would be valid if the "control" was nonreactive -- e.g. a
lamp-post that would always be a lamp-post regardless of the hypothesis,
regardless of skill, and even with a wig would still be a lamp-post.

That doesn't apply in this case because the experimenter is potentially
reactive.

And probably increasingly paranoid as the experiment went on! I'd put a
smiley there, but it's really a pretty sad thing. This guy is now making the
assumption that, if you ride a bike on the road, you're going to get hit by
a car.

--Mike-- Chain Reaction Bicycles
www.ChainReactionBicycles.com

"Mike Kruger" <[email hidden]> wrote in message
news:[email hidden]...

Quoted message said:

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

Quoted message said:

Chris Neary said:

Wow, so we have really great data applicable to ONE person.

The problem isn't that ONE person collected the data - the problem is
that
ONE person was the experiment from which the data was collected.

No. We have data regarding 2,300 vehicles passing a control. The
passing vehicles were the experiment, the cyclist was the control. The
control conditions were changed to see what the experiment participants
(although they didn't know they were participating) would do.

Your argument would be valid if the "control" was nonreactive -- e.g. a
lamp-post that would always be a lamp-post regardless of the hypothesis,
regardless of skill, and even with a wig would still be a lamp-post.

That doesn't apply in this case because the experimenter is potentially
reactive.

--
"President Washington, President Lincoln, President Wilson, President
Roosevelt have all authorized electronic surveillance on a far broader
scale." --Attorney General Alberto Gonzalez

[email hidden] wrote:
....

Quoted message said:

As http://www.zzapp.org/rileygea/itsa/helmet2.htm points out,
information on something as simple as "Was the cyclist wearing a
helmet" is _massively_ unreliable on a national scale. We have no
hopes of sifting existing information to learn even that, let alone
anything more.

I agree completely. The data isn't there.

Frank, how will you claim that this guy's ridiculous
'study' has merit without also coming to the conclusion
that cycling is ungodly dangerous?

At his rate of collision, I would have been struck by
vehicles over 2,500 times so far.

Robert

Mike Kruger said:

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

Quoted message said:

No. We have data regarding 2,300 vehicles passing a control. The
passing vehicles were the experiment, the cyclist was the control. The
control conditions were changed to see what the experiment participants
(although they didn't know they were participating) would do.

Your argument would be valid if the "control" was nonreactive -- e.g. a
lamp-post that would always be a lamp-post regardless of the hypothesis,
regardless of skill, and even with a wig would still be a lamp-post.

That doesn't apply in this case because the experimenter is potentially
reactive.

I suppose by "reactive" you mean that the cyclist would be adjusting
his lane position based on the passing distance of the vehicles.

I see several problems with this logic. First, any immediate reaction
would likely occur after the vehicle has already passed, thus negating
the effects of the reaction on the distance measurement. Second, you
assume that the cyclist does not have a standard method for measuring
distance to road edge and maintaining proper lane position. Third, the
assumption of long-term conditioning would suggest that there would be
no cyclists on the road as we would all eventually be bullied off the
road by passing motorists.

What you are suggesting is that the experiment is impossible to conduct
because every cyclist will be "reactive" to some extent. I'd rather see
some data from a single cyclist than a gaggle of cyclists because it is
easier to control all of the potential variables.

-Buck

Quoted message said:

[email hidden] wrote:
...

Quoted message said:

As http://www.zzapp.org/rileygea/itsa/helmet2.htm points out,
information on something as simple as "Was the cyclist wearing a
helmet" is _massively_ unreliable on a national scale. We have no
hopes of sifting existing information to learn even that, let alone
anything more.

I agree completely. The data isn't there.

Frank, how will you claim that this guy's ridiculous
'study' has merit without also coming to the conclusion
that cycling is ungodly dangerous?

At his rate of collision, I would have been struck by
vehicles over 2,500 times so far.

Everyone is focused on the collisions when they should be focused on
the study. I actually bothered to do some digging and discovered that
he has a great deal of cycling experience. Here's what he posted on his
website: http://www.drianwalker.com/cycling.html
Now, how many of the pundits around here have cycled solo across
Europe?

I do wonder what happened in those incidents. But I won't assume that
both of these events were his fault or that he could have done
something to prevent either one. They are probably just a run of bad
luck. Strange things happen in quick succession sometimes.

Most research comes from a passion about a subject. Dr. Walker appears
to be someone with a passion for cycling. He designed an experiment
which had enough scientific rigor to warrant funding. The work was
placed under scientific review and passed for publication. He had a
couple of incidents while conducting the experiments. He had the
fortitude to report those incidents as he should.

When it is published, if Frank and Jobst and whomever else around here
with a background in scientific research start poking holes in the
methodology, then I'll consider there may be a problem with the study
design.

-Buck

Quoted message said:

What you are suggesting is that the experiment is impossible to conduct
because every cyclist will be "reactive" to some extent. I'd rather see
some data from a single cyclist than a gaggle of cyclists because it is
easier to control all of the potential variables.

Guess it depends upon what you want to study. We know very well how things
affect this one particular cyclist. We know little, if anything, about how
things affect your or I. As has been pointed out many, many times, this
particular cyclist is a statistical anomaly. Getting hit twice in such a
short period of time, and both times apparently while wearing a helmet...

One thing that would be interesting to know is whether he actually looked at
his data more-or-less in real time, or checked everything only after the
experiment was concluded. That could be a *huge* factor.

In any event, we might agree that a better study would involve many cyclists
involved in many passing situations each. And in no event should the
participants be aware of the measurements at any time during the study.

--Mike-- Chain Reaction Bicycles
www.ChainReactionBicycles.com

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

Quoted message said:

Mike Kruger said:

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

Quoted message said:

No. We have data regarding 2,300 vehicles passing a control. The
passing vehicles were the experiment, the cyclist was the control. The
control conditions were changed to see what the experiment participants
(although they didn't know they were participating) would do.

Your argument would be valid if the "control" was nonreactive -- e.g. a
lamp-post that would always be a lamp-post regardless of the hypothesis,
regardless of skill, and even with a wig would still be a lamp-post.

That doesn't apply in this case because the experimenter is potentially
reactive.

I suppose by "reactive" you mean that the cyclist would be adjusting
his lane position based on the passing distance of the vehicles.

I see several problems with this logic. First, any immediate reaction
would likely occur after the vehicle has already passed, thus negating
the effects of the reaction on the distance measurement. Second, you
assume that the cyclist does not have a standard method for measuring
distance to road edge and maintaining proper lane position. Third, the
assumption of long-term conditioning would suggest that there would be
no cyclists on the road as we would all eventually be bullied off the
road by passing motorists.

What you are suggesting is that the experiment is impossible to conduct
because every cyclist will be "reactive" to some extent. I'd rather see
some data from a single cyclist than a gaggle of cyclists because it is
easier to control all of the potential variables.

-Buck

I had a few email exchanges with the author several days ago.

Here's one:

Me: Since you were struck twice in a short period, what is your
explanation? What were the road, traffic, driver, and your conditions?
What did the drivers say if anything?

Response: I was struck by a bus and a truck - both got too close
passing. In normal circumstances I would have been able to move away
when I felt them getting close, but the study being what it was, I
couldn't move every time I felt a vehicle getting close as not only
would I lose those data, they were the most interesting data!

He knows I've written a bunch of papers at www.humantransport.org,
including collision studies. I think his answer to my question was a
dodge. However, if he thinks he could have moved away, they must have
been very very slow. I think it's possible he was passing on the right,
or he's making it up entirely. Definitely aggrandizing.

Wayne