Bicycle Accidents - Statistical Analysis

[citydreams]

Is there some way of meaningfully relating cyclist injuries/fatalities to health benefits of cycling?

For example, if I were to cycle to work for a year, would you be able to predict an increase in my life expectancy/general quality of life and to what extent would that be balanced out by the increased chance of death/long term injury?

There are various measures used.. and quite a lot of argument about whether there's double counting.. under counting or over counting.. For example, if you cycle then you're not walking as far.. but the life expectancy numbers are often based on people who don't do any exercise..

The cycling england report is very interesting though... http://www.dft.gov.uk/cyclingengland/health-fitness/health-benefits-of-cycling/

If you play with the WHO-HEAT too that I posted above, and put in your age, fitness levels and so on.. it should tell you how much longer (on average) you'd live for.

 

[two sheds]

I'm also just finishing an MSc dissertation on the cost-benefit analysis of walking and cycling tracks. Really struggling with it at the moment.

Have you graphs showing pilot studies on how walking/cycling tracks reduce the number of accidents? Same sort of approach would you'd think apply (and remember that i speak here from blissful ignorance ).

They do similar with predicting energy consumption - if you can see a difference before/after any improvements from the process behaviour chart you can draw cusum to show the cumulative sum of the difference. If you know the cost of implementing the measures and (for example) the savings in hospital bills and work lost etc you should be able to do a comparison.

Then of course, you've got the immeasurable costs for the individuals of not having to go through several months of pain after an accident.

 

[citydreams]

Have you graphs showing pilot studies on how walking/cycling tracks reduce the number of accidents? Same sort of approach would you'd think apply (and remember that i speak here from blissful ignorance ). .

Haven't got any real GIS data for the cycle routes that's in a useable format...

Could have a look at Vauxhall Bridge to see if there is any change in the area there. Though I'm sure the Centre of Cycling Excellence at TfL would have done that already.. Still worth a look I reckon

Not sure I'd use the same approach there.. as I'd need a dummy value for the building works associated with putting in the scheme.. probably use a simple linear model:
accidents = cycle.flow + bus.flow + car.flow + cycle_lane_%_completed + rain/temp

though a time series would be good to look for sudden changes from the control..

 

[citydreams]

I've got 450 walking + cycling accidents around Vauxhall Bridge over the 1994 - 2008 range.

Just need to find out what cycling measures were put in place over the building works for the new bus station.

 

[citydreams]

Yep, seasonality would need to be taken into account, don't know about sine and cosine (i refer you back to my opening sentence .

The idea comes from electronics.. Every curve is in fact made up of lots of other curves all stacked on top of each other.. So any curve can me made up of sine and cosines of varying phases, amplitutes and pitches.. By playing with them, you can work out seasonality, and look for underlying trends within your dataset..

 

[Maggot]

There was a report in the news (last year IIRC) which said that female cyclists were more likely to be killed than male ones due to them not jumping red lights and being hit by vehicles turning left. I heard that this was rubbish cos it was based on a tiny sample, but have not had any proof either way. What's the story?

And good luck with your MSc.

 

[two sheds]

Not sure I'd use the same approach there.. as I'd need a dummy value for the building works associated with putting in the scheme..

that would just be a one-off cost on the 'costs' side, though, plus any reduction in speeds/increase in accidents from having two lanes rather than three for the rest of the traffic for example?

probably use a simple linear model: accidents = cycle.flow + bus.flow + car.flow + cycle_lane_%_completed + rain/temp

Interesting yep, also dark/light? I've also heard that half-completed cycle lanes can be worse than none at all because a danger point is where the cyclists funnel into the main traffic again.

though a time series would be good to look for sudden changes from the control..

For process behaviour charts they also recommend plotting over different timescales, so weekly ones might show the direct effect of people getting pissed up at christmas/new year, daily ones might show effects of work/weekend days, and hourly ones might show differences from streetlighting.

 

[citydreams]

There was a report in the news (last year IIRC) which said that female cyclists were more likely to be killed than male ones due to them not jumping red lights and being hit by vehicles turning left. I heard that this was rubbish cos it was based on a tiny sample, but have not had any proof either way. What's the story?

http://www.movingtargetzine.com/article/red-lights-and-media-spin

The main problem is lack of data really.. I tried to get TfL to commission some more research but they were'nt having it.. though I think that's going to change due to the UK statistics authority recently giving the DfT a telling off

And good luck with your MSc

thanks I've passed the exams already, but would be nice to ace the dissertation!

 

[citydreams]

that would just be a one-off cost on the 'costs' side, though, plus any reduction in speeds/increase in accidents from having two lanes rather than three for the rest of the traffic for example?

Aye.. though flow is really a good indication of the speed anyway on routes like Vauxhall bridge as there are capacity issues

[For process behaviour charts they also recommend plotting over different timescales, so weekly ones might show the direct effect of people getting pissed up at christmas/new year, daily ones might show effects of work/weekend days, and hourly ones might show differences from streetlighting.

I reckon you might enojy this.... http://www.stat.pitt.edu/stoffer/tsa2/R_time_series_quick_fix.htm

 

[toblerone3]

>>>toblerone3

here you go:
http://www.euro.who.int/transport/policy/20081219_1

I've got a copy if you want to have a look..

Thanks for your thoughts. This is exactly what I'm looking into.

Namely one of the default parameters in the WHO's "Health Economic Assessment Tool" that assumes that if you build a new cycle track and measure the amount of cycling exercise that occurs on it you can assume that 50% of that exercise is additional to what new and existing cyclists would have done anyway on other routes.

And also including the phenomena of exercise displacement where people who start cycling more might feel the need to take part in other forms of excercise such as sport or the gym less.

This is a key input into working out the health benefits of walking and cycling infrastructure.

Both health benefits and accident trends are key benefits in the NATA framework.

 

[two sheds]

The idea comes from electronics.. Every curve is in fact made up of lots of other curves all stacked on top of each other.. So any curve can me made up of sine and cosines of varying phases, amplitutes and pitches.. By playing with them, you can work out seasonality, and look for underlying trends within your dataset..

Interesting, ta. The idea is to allow for the trends like that, to adjust the readings so you're actually working from a flat graph?

One thing they do with finding waste in buildings is to model the actual processes, too, so separating out the different uses of the energy - calculating how much for lights, heating, computers, etc etc. should be consumed. That should give you the average value you see from the actual graphs of consumption. It doesn't really help in finding waste, but does help when you come to invest in the actual energy saving measures or upgrade the building/equipment.

Equivalent here I presume would be to separate out the different types of accidents (20/80 Pareto style) and how likely they are. Not sure whether it would be relevant, but I can imagine it would have useful information if you're trying to design a traffic control type system. Mind you, looking at that again it's just a pie chart.

For quality improvement in integrated circuits and other manufacturing they often use Ishikawa diagrams to try to relate causes and effects. Relevant you'd think because if anybody knows about quality improvement, it's integrated circuit manufacturers.

 

[citydreams]

- Do most cyclist accidents happen in moving traffic, or starting from a halt at a junction, or coming to a halt at a junction?

no real data for that.. only thing I can say is where abouts accident happened.. on a roundabout; 50 metres from junction; cross-roads &c..

we're lucky to have any data at all!

- What kind of junctions are most dangerous? Do ASLs seem to have any impact?

the closer you get to the centre of London, the more dangerous it gets.. sorry, I know that's a bit flippant.. still crunching the numbers on that one.. tried running a latent cluster analysis but I didn't get any joy.. currnently running a CHAID..

though don't have any data on ASLs to go on.

How many cyclist / pedestrian casualties are caused by cars, HGVs, buses, motorcyclists?

will split it out and post results later (though I would say all of them )

- What factors were involved when there was a motor vehicle at fault? E.g. speeding, careless driving, etc.

too many to list! can you narrow the question?

- How many cyclist accidents happen after going the wrong way on a one-way street or breaking red lights?

no real data to answer this question

- Is Inner London safer than outer London for cyclists?

I think I've come to the conclusion that safer is a very subjective term!

Safer in terms of the chance of being in an accident?

Or safer in terms of severity of injury?

Or safer in terms of relative conditions improving compared to inner London. ?

Will post my findings once I've written up.

- Mightn't some quite dangerous roads 'look safe' because pedestrians and cyclists avoid them altogether?

...or that the accidents weren't reported by the police.. that's the only way we can count them.

Sorry, I recognise that's quite a lot, and some of it probably isn't answerable, but you did ask!
.

keep em coming Will give you some proper answers once I've put my findings in order.

 

[two sheds]

I reckon you might enojy this.... http://www.stat.pitt.edu/stoffer/tsa2/R_time_series_quick_fix.htm

Ta - bookmarked will have a peruse.

 

[citydreams]

Thanks for your thoughts. This is exactly what I'm looking into.

Sounds like you've got it sorted!

Let me know if I can help at all.

 

[two sheds]

The other thing that comes to mind is the accuracy of the statistics - as i understand it, it's no use trying to analyse unemployment and possibly crime figures because they keep changing the way they collect the fucking figures.

 

[citydreams]

The other thing that comes to mind is the accuracy of the statistics - as i understand it, it's no use trying to analyse unemployment and possibly crime figures because they keep changing the way they collect the fucking figures.

Aye.. i found that recently analysing air quality in London... Someone changed the Heathrow weather station and didn't tell!

 

[citydreams]

- How many cyclist accidents happen after going the wrong way on a one-way street or breaking red lights?
-.

might be possible to infer this from the RNPR Traffic Note 8.. "Proportion of Cyclists Who Violate Red Lights in London"... hmmm...

 

[two sheds]

Aye.. i found that recently analysing air quality in London... Someone changed the Heathrow weather station and didn't tell!

Mind you on reflection, analysing the government's unemployment figures might actually show clearly the effectiveness of their 'back to work' efforts. You'd have the unemployment graph going upwards that suddenly flattened out and you could say 'yes that's where thatcher moved people onto sickness benefits' and later 'that's where they tightened up the rules for when you qualify for benefit'.

 

[nick h.]

I'd like to know whether roads with higher speed limits are more dangerous for cyclists. They feel scary at first but they tend not to have parked cars or random pedestrians. And they're wider. With fewer junctions. When I was in a cycling club we all trained on dual carriageways. The noise wasn't pleasant but it seemed pretty low risk.

I'd also like to know how much difference hi-viz clothes make.

 

[yojimbo]

Thanks for the answers, citydreams. I guess with the question about factors behind crashes I'm mainly interested in whether enforcement of speed limits would make a big difference, although I recognise that the details of 'cause of accident' you get are probably quite problematic - e.g. is 'speeding' down to lack of enforcement or road design or poor signposting, etc. Also, you're right that 'safer' can mean all sorts of things, and any one indicator is probably going to be incomplete.

Oh and one more question - does your data separate out taxis from cars in general?

Best of luck with the rest of the work, I'm looking forward to seeing the results.

 

[citydreams]

..

 

[toblerone3]

What is STAT19 and STATS19?

 

[Paulie Tandoori]

so...... are things getting any better?

good graphs btw

 

[Miss-Shelf]

sounds really interesting

good luck (I am finishing up my dissertation(not on cycling) - did I say finishing? and know how it drags)

look forward to hearing your conclusions

I'd be interested to know how many vehicle drivers were found culpable - not sure that'd be possible from your data