[Mungecrundle]

https://www.bbc.co.uk/news/av/technology-46790235/crash-testing-the-air-bag-for-cyclists

Go!

Topic

[Rich_cb]

alansmurphy wrote:

alansmurphy wrote:

The part I disagree with Rich is using any data from a group that confirms that wearing a hi viz jacket prevents you falling off your bike!

 

The recognised methodology they should be using is deleting all the data.

 

And 2/3rds wearing a jacket – being aware that they aren’t all early/late riders – c’mon open your eyes…

You claim to have some understanding of research methodology.

The above post contradicts that somewhat.

Bias is unavoidable in unblinded studies.

In this study that bias manifests itself as an undereporting of accidents in the yellow jacket group.

The authors identified the bias and controlled for it using a recognised published technique.

You might not agree with that technique but unless your criticism amounts to more than just calling it ‘bollocks’ I’m afraid it doesn’t carry much weight and, ironically, is probably just a reflection of your own bias.

[alansmurphy]

The part I disagree with Rich

The part I disagree with Rich is using any data from a group that confirms that wearing a hi viz jacket prevents you falling off your bike!

 

The recognised methodology they should be using is deleting all the data.

 

And 2/3rds wearing a jacket – being aware that they aren’t all early/late riders – c’mon open your eyes…

[FlyingPenguin]

alansmurphy wrote:

Further into the study it shows that there are seasonal variations on jacket use, November had the highest proportion wearing (well it is colder and darker) and July the lowest at only 67%. What? Are you seriously suggesting that in the middle of summer 2 in 3 riders were wearing a jacket?!

 

Surely the sort of nodder that dresses like a traffic island year round is exactly the sort of person that would agree to be part of a study like this!

[Rich_cb]

alansmurphy wrote:

alansmurphy wrote:

I like that you can use ‘Control C’ and ‘Control V’ but their adjustment is frankly bollocks. It is now you that is looking for something to support your predetermined viewpoint and ignoring what goes against it.

 

If nearly half of your sample are talking shit about the incidents then you’re in trouble. If 2/3 are claiming they wear the jacket in August then you’re in trouble when analysing the data. Simple really.

 

Again, like the helmet debate, I wear hi viz. I’m about to leave work and put on Northwave reflective shoe covers, a high viz flashing strap on each leg, an Altura nightvision reflective jacket, have reflective detailing on the galibier gloves. None of this will prevent me falling off my bike, it simply makes me feel more likely to be seen or gives the tw@t that knocks me off many more items to try and throw over a hedge for his SMIDSY defence.

The authors are using a recognised technique to control for the inevitable bias that all research encounters in unblinded studies.

You claim that the adjustment is ‘bollocks’. Care to elaborate? Which part do you disagree with?

Also, according to Wikipedia, the average July temp in Denmark is a balmy 17 degrees. Mornings and evenings are likely to be cooler so it’s not inconceivable that people might wear a jacket.

[alansmurphy]

I like that you can use

I like that you can use ‘Control C’ and ‘Control V’ but their adjustment is frankly bollocks. It is now you that is looking for something to support your predetermined viewpoint and ignoring what goes against it.

 

If nearly half of your sample are talking shit about the incidents then you’re in trouble. If 2/3 are claiming they wear the jacket in August then you’re in trouble when analysing the data. Simple really.

 

Again, like the helmet debate, I wear hi viz. I’m about to leave work and put on Northwave reflective shoe covers, a high viz flashing strap on each leg, an Altura nightvision reflective jacket, have reflective detailing on the galibier gloves. None of this will prevent me falling off my bike, it simply makes me feel more likely to be seen or gives the tw@t that knocks me off many more items to try and throw over a hedge for his SMIDSY defence.

[Rich_cb]

alansmurphy wrote:

alansmurphy wrote:

Thanks, I have a basic grap of statistics and research methodology.

 

What I’m telling you is the methodology and the outcomes demonstrates a huge bias hence the rest of the outcomes will suffer from this.

 

Unless you can explain how wearing a high viz jacket reduces the ‘nobody else involved’ incidents by nearly a half, you cannot expect anyone to take the rest of the results seriously…

If only the authors had noticed that.

They may have devoted an entire section to explaining how they controlled for the inevitable bias.

It may have gone something like this:

“The bicycle jacket is expected to influence the number of multiparty PIAs, whereas the number of single PIAs is not likely to be affected by jacket wearing. Therefore, any difference between the test and control groups indicates a potential reporting bias between the groups, which may occur because the study is non-blinded, and the participants know whether they belong to the test or control group. We adjusted for potential report bias as suggested by Madsen et al. (2013) by correcting the risk of multiparty PIAs in the control group, using a correction factor equal to the estimated ARR for single PIAs as a general correctional factor

The index j refers to type of accident (e.g. winter accidents) and the index C means “Control group”. Specifically,
with data from Table 3, Table 4.

This correctional factor was multiplied by the reported number of PIAs in the control group, thus reducing the number of PIAs in the control group caused by the likely underreporting of PIAs in the treatment group (Madsen et al., 2013). The corrected ARs for the control group were estimated in the following way:

In practice, the corrected analysis was based on an estimate of a corrected ARR given by the relative difference between the multiparty ARR and the single ARR:

As the ARRs are likely to be skewed, the standard errors are usually estimated through a transformation to the logarithmic scale. The standard error of the corrected ARR of multiparty PIAs on the logarithmic scale was given by:

And the formula of the 95% confidence interval on the logarithmic scale:

The 95% limits were then back-transformed. In practice, the standard error on a logarithmic scale can be estimated in the following way:
where X denotes the number of PIAs, i is either “multiparty” or “single” and T denotes the test group and C the control group.”

[alansmurphy]

Thanks, I have a basic grap

Thanks, I have a basic grap of statistics and research methodology.

 

What I’m telling you is the methodology and the outcomes demonstrates a huge bias hence the rest of the outcomes will suffer from this.

 

Unless you can explain how wearing a high viz jacket reduces the ‘nobody else involved’ incidents by nearly a half, you cannot expect anyone to take the rest of the results seriously…

[Rich_cb]

alansmurphy wrote:

alansmurphy wrote:

So just putting a high viz jacket on makes you only around half as likely to fall off your bike (PIAs seemingly being accidents not involving another road user).

 

So if i put it on, take it off, put it on, take it off and put it on again then I’m invincible!

 

Brilliant!

It is impossible to blind the participants in this particular study.

Bias is inevitable in unblinded studies.

The authors identified and controlled for the unavoidable bias.

The analysis was carried out using ‘intention to treat’ hence the inclusion of those who chose not to wear the jacket.

https://en.m.wikipedia.org/wiki/Intention-to-treat_analysis

[ClubSmed]

alansmurphy wrote:

Further into the study it shows that there are seasonal variations on jacket use, November had the highest proportion wearing (well it is colder and darker) and July the lowest at only 67%. What? Are you seriously suggesting that in the middle of summer 2 in 3 riders were wearing a jacket?!

Must have been cycling in Scotland 

[alansmurphy]

Also, “This principle

Also, “This principle compares all participants in the randomised groups, even if some participants in the test group did not wear the jacket, and prevents confounding bias”

 

Well this won’t work as presumably you’d be underestimating the effectiveness of the yellow jacket.

 

Further into the study it shows that there are seasonal variations on jacket use, November had the highest proportion wearing (well it is colder and darker) and July the lowest at only 67%. What? Are you seriously suggesting that in the middle of summer 2 in 3 riders were wearing a jacket?!

[alansmurphy]

stomec wrote:

Do you have a problem with the study design/methodology, does it conflict with previous research you are aware of, or is just that the result is not in keeping with what you imagined it to be in your head?

 

whatever research wrote:

The accident rate (AR) (accidents per person month) for personal injury accidents (PIAs) for the test group was 47% lower than that of the control group. 

 

So just putting a high viz jacket on makes you only around half as likely to fall off your bike (PIAs seemingly being accidents not involving another road user).

 

So if i put it on, take it off, put it on, take it off and put it on again then I’m invincible!

 

Brilliant!

[hawkinspeter]

stomec wrote:

HawkinsPeter wrote:

Wow – that research shows an unreasonably large effect from the hi-viz – 38% reduction.

This is an interesting comment.  Do you have a problem with the study design/methodology, does it conflict with previous research you are aware of, or is just that the result is not in keeping with what you imagined it to be in your head?

It just conflicts with what I’d expect. I would have guessed that hi-viz would maybe have a 10% effect at best. As I currently cycle a lot at night, I prefer to go with reflectives (especially the 3M high reflective stickers).

I haven’t drilled down into the study design/methodology though I did spot that it wasn’t a blinded study (presumably they kept falling off when they were blinded).

[stomec]

HawkinsPeter wrote:

Wow – that research shows an unreasonably large effect from the hi-viz – 38% reduction.

This is an interesting comment.  Do you have a problem with the study design/methodology, does it conflict with previous research you are aware of, or is just that the result is not in keeping with what you imagined it to be in your head?

[ClubSmed]

HawkinsPeter wrote:

[quote=Rich_cb] Research: https://www.sciencedirect.com/science/article/pii/S0925753517313528%5B/quote%5D

Wow – that research shows an unreasonably large effect from the hi-viz – 38% reduction.

That is a very large effect! If you take this figure and add the data from the Daytime Conspicuity Benefits of Fluorescent Bicyclist Apparel (https://journals.sagepub.com/doi/abs/10.1177/1541931213601954) that would mean that wearing hi-vis leggings makes you pretty much invulnerable. Why are there not many of these on the market?

[hawkinspeter]

Rich_cb wrote:

Rich_cb wrote:

[quote=Simon E]But saying “not always effective” or “will only help in certain crash scenarios” aren’t great sales lines. How far down the road of “if it saves one life” should we go? Full leathers and a motorcycle helmet? That’s going be better than 0.3mm of lycra and a 200g hat full of holes, isn’t it?

 

I was posing the question in a provocative way but I feel it’s a valid point: why do people wearing hi-viz still get hit? And do riders in black get knocked off more than those wearing brighter colours? If so then are the clothing brands irresponsible for selling it?

 

Do you have links to any of this research? I’m prepared to revise my opinion if it is convincing.

 

You think the airbag is a good idea, apparently without that kind of evidence. So it’s fine for others to buy stuff based on marketing hype but I can’t go with a scepticism resulting from decades of experience, a ‘gut feeling’ based on my research and experience?

Will you wear an airbag once they become available?

For me it’s a balance between safety and comfort. It’s been pretty cold recently so I’ve been wearing a gilet, do I care what colour my gilet is? Not really. So my comfort is identical but my safety is improved. Full leathers and a motorcycle helmet would probably improve safety in a collision but would be very uncomfortable so not really an option. People in hi viz still get hit because the majority of all collisions are not preventable by the cyclist. All we can do is focus on the collisions we can prevent through good equipment and good road craft. I think equipment that only protects the cyclist, eg helmet, high visibility, should not be compulsory and as such while I wouldn’t personally go out in the dark dressed entirely in black I think others should be free to do so. I think the airbag is a good idea without research as it’s literally just been invented so gut feeling is all I have to go on, I probably won’t buy one now but may buy a future iteration once the evidence is available. Research: https://www.sciencedirect.com/science/article/pii/S0925753517313528%5B/quote%5D

Wow – that research shows an unreasonably large effect from the hi-viz – 38% reduction.

You’re right about the safety/comfort balance (though maybe safety/comfort/price might be more accurate in this instance), but I’m really not convinced by torso protection on a bike. It would probably make more sense for me to wrap loads of bubble wrap around the street furniture alongside my regular routes (cheaper too).

[Author]

Replies