What is it with women cyclists?

rhext

All these posters and no statisticians!

Someone must know how to work out whether these stats show women significantly more likely to be sideswiped by a HGV than men.

If I could remember any of my A level maths, I could probably do it myself. Unfortunately, over 20 years of work and beer have taken their toll!

roger_merriman

rhext wrote:

All these posters and no statisticians!

Someone must know how to work out whether these stats show women significantly more likely to be sideswiped by a HGV than men.

If I could remember any of my A level maths, I could probably do it myself. Unfortunately, over 20 years of work and beer have taken their toll!

no one would touch it with a barge pole thats why, the sample size is way too small, even the total deaths per years are too small. the numbers are just too small to use as stats.

rhext

roger merriman wrote:

rhext wrote:

All these posters and no statisticians!

Someone must know how to work out whether these stats show women significantly more likely to be sideswiped by a HGV than men.

If I could remember any of my A level maths, I could probably do it myself. Unfortunately, over 20 years of work and beer have taken their toll!

no one would touch it with a barge pole thats why, the sample size is way too small, even the total deaths per years are too small. the numbers are just too small to use as stats.

Disagree, there are statistical techniques for dealing with small sample sizes (seem to remember something about a 'T' distribution from my A level days), particularly when those samples are drawn from a very large population. Such techniques are often used to identify (for example) whether disease clusters are significant based on very small numbers of cases.

always_tyred

rhext wrote:

roger merriman wrote:

rhext wrote:

All these posters and no statisticians!

Someone must know how to work out whether these stats show women significantly more likely to be sideswiped by a HGV than men.

If I could remember any of my A level maths, I could probably do it myself. Unfortunately, over 20 years of work and beer have taken their toll!

no one would touch it with a barge pole thats why, the sample size is way too small, even the total deaths per years are too small. the numbers are just too small to use as stats.

Disagree, there are statistical techniques for dealing with small sample sizes (seem to remember something about a 'T' distribution from my A level days), particularly when those samples are drawn from a very large population. Such techniques are often used to identify (for example) whether disease clusters are significant based on very small numbers of cases.

Yes, but these are still methods of statistically polishing a tur.d.

rhext

Always Tyred wrote:

rhext wrote:

roger merriman wrote:

rhext wrote:

All these posters and no statisticians!

Someone must know how to work out whether these stats show women significantly more likely to be sideswiped by a HGV than men.

If I could remember any of my A level maths, I could probably do it myself. Unfortunately, over 20 years of work and beer have taken their toll!

no one would touch it with a barge pole thats why, the sample size is way too small, even the total deaths per years are too small. the numbers are just too small to use as stats.

Disagree, there are statistical techniques for dealing with small sample sizes (seem to remember something about a 'T' distribution from my A level days), particularly when those samples are drawn from a very large population. Such techniques are often used to identify (for example) whether disease clusters are significant based on very small numbers of cases.

Yes, but these are still methods of statistically polishing a tur.d.

rhext

Always Tyred wrote:

rhext wrote:

roger merriman wrote:

rhext wrote:

All these posters and no statisticians!

Someone must know how to work out whether these stats show women significantly more likely to be sideswiped by a HGV than men.

If I could remember any of my A level maths, I could probably do it myself. Unfortunately, over 20 years of work and beer have taken their toll!

no one would touch it with a barge pole thats why, the sample size is way too small, even the total deaths per years are too small. the numbers are just too small to use as stats.

Disagree, there are statistical techniques for dealing with small sample sizes (seem to remember something about a 'T' distribution from my A level days), particularly when those samples are drawn from a very large population. Such techniques are often used to identify (for example) whether disease clusters are significant based on very small numbers of cases.

Yes, but these are still methods of statistically polishing a tur.d.

But that's the beauty of it, they're not. They provide a probability of the difference being statistically significant. Anyway, never one to back down from a challenge, it's off to Wikipedia.

If we start with a null hypothesis which says that when someone is killed by a lorry it is just as likely to be a man as it is a woman, take a sample size of 7 of which 6 were women, assuming I've done the maths right, you've got a 12.5% probability of the null hypothesis being right (two-tailed binomial distribution significance test). The 'usual' test for significance would be 5%, so we can conclude that these stats don't show a significantly increased risk for women.

Incidentally, if it were 7 out of 7, however, you're only on a 1.5% probability of the 'equally likely' hypothesis to be correct which would be significant at a 5% confidence level. So it's not necessarily the sample size which drives the assessment.

biondino

Well, it is the sample size. Let's say 7 women were killed, and you're claiming it is statistically relevant - and then a man gets himself thoughtlessly killed, and despite having a larger sample size the statistics become irrelevant.

It really, really isn't a big enough sample, even if we had perfect knowledge of the circumstances regarding every death.

DonDaddyD

The sample size is a too small to be truly representative. The ratio of left sided accidents involving women, in recent times, compared to men might be high (6 accidents 7 being women). But as a percentage of the total number of accidents involving cyclists compared to the number of cyclists nationally is not a truly representative example.

Its like saying 7 people in London caught swine flu, of those seven, six of them were women! Lets make it a pandemic! The ordinary flu, has infected and killed more in a shorter span of time... - or so I've been briefed at work in a hospital.

However in and of itself the number of women finding themselves in left sided situations is alarming... that said I thought this was a piece of genius:

We don't know whether the HGV driver did or did not target the cyclist, we don't know whether they did or did not see the cyclist, we don't know if the driver was or wasn't reacting because it was a women and we don't know whether it was or wasn't because of the cyclists endangered themselves (namely the supposed female tendency to ride up the left hand side).

So until a clear and definitive reason can be given as to why accidents of this nature always seem to happen to women, we have to speculate that its a combination of all possible and reasonable outcomes and therefore advise, recommend best practice and act accordingly taking all accounts into consideration.

RedJohn

rhext wrote:

If we start with a null hypothesis which says that when someone is killed by a lorry it is just as likely to be a man as it is a woman, take a sample size of 7 of which 6 were women, assuming I've done the maths right, you've got a 12.5% probability of the null hypothesis being right (two-tailed binomial distribution significance test). The 'usual' test for significance would be 5%, so we can conclude that these stats don't show a significantly increased risk for women.

Incidentally, if it were 7 out of 7, however, you're only on a 1.5% probability of the 'equally likely' hypothesis to be correct which would be significant at a 5% confidence level. So it's not necessarily the sample size which drives the assessment.

But in fact, given that there are considerably more male cyclists than there are female ones, let's guess 4 x as many, your null hypothesis should in fact be that when someone is killed by a lorry it is 4x as likely that they are a man. So when in fact 6 out of 7 are women, the probability of the null hypothesis being right is considerably smaller.

I'll let you figure out how small though :-)

rhext

biondino wrote:

Well, it is the sample size. Let's say 7 women were killed, and you're claiming it is statistically relevant - and then a man gets himself thoughtlessly killed, and despite having a larger sample size the statistics become irrelevant.

It really, really isn't a big enough sample, even if we had perfect knowledge of the circumstances regarding every death.

Nothing is certain in statistics....all you can do it state confidence levels and relate them to somewhat arbitrary 'significance' tests. According to the 'usual' 95% confidence test, 6 out of 7 doesn't allow you to say 'I'm confident that any given HGV death is more likely to be female than male'. 7 out of 7 would allow you to make that statement, but there's still a probability (albeit fairly small) that your next data point might reverse that position.

But even this analysis is actually answering the wrong question, the 'right' one being something like 'if a woman cycles for one mile, is she more likely to get killed by a truck than a man cycling the same mile'. In order to work that out, we'd need to know the ratio of 'miles cycled by men' to 'miles cycled by women'. And that's crucial in working out the significance of the stats. If (for example) twice as many miles were cycled by men as by women, but even given this 6 out of 7 HGB sideswipes were women, then that would be highly significant. Again (assuming my vaguely remembered high school maths is roughly right), a null hypothesis which says 'women and men are equally likely to get swiped in a given mile' is over 99% likely to be wrong under those conditions.

ellieb's stats would also tend to point to the same conclusion, incidentally, and they've got a much larger sample size!

tarquin_foxglove

A discussion about this is on women's hour on R4 now with 4 commuters from Bristol.

tarquin_foxglove

Just ditched the commuters & have Police and someone else on now.

tarquin_foxglove

Police blaming the cyclists for wobbling.

Cafewanda

Idiotic police :roll:

dilemna

http://www.bbc.co.uk/programmes/b00lfdlr

Other guest is Cynthia Barlow of Roadpeace whose own daughter was killed by an HGV.

Yep plod Jo Oakley is deffers ignorant and clearly not a cyclist making such ill informed remarks. Those women cyclists that have been killed by HGVs have been very competent and experienced re the latest Catriona at the Oval tube station. No mention of recent deaths or those at Blackfriars Bridge earlier in the year.

Unfortunately not enough time or weight was given to the issue, especially when a subsequent article was about female nakedness in art! A very odd selection of subject matters for the programme.

always_tyred

rhext wrote:

Always Tyred wrote:

rhext wrote:

roger merriman wrote:

rhext wrote:

All these posters and no statisticians!

Someone must know how to work out whether these stats show women significantly more likely to be sideswiped by a HGV than men.

If I could remember any of my A level maths, I could probably do it myself. Unfortunately, over 20 years of work and beer have taken their toll!

no one would touch it with a barge pole thats why, the sample size is way too small, even the total deaths per years are too small. the numbers are just too small to use as stats.

Disagree, there are statistical techniques for dealing with small sample sizes (seem to remember something about a 'T' distribution from my A level days), particularly when those samples are drawn from a very large population. Such techniques are often used to identify (for example) whether disease clusters are significant based on very small numbers of cases.

Yes, but these are still methods of statistically polishing a tur.d.

But that's the beauty of it, they're not. They provide a probability of the difference being statistically significant. Anyway, never one to back down from a challenge, it's off to Wikipedia.

If we start with a null hypothesis which says that when someone is killed by a lorry it is just as likely to be a man as it is a woman, take a sample size of 7 of which 6 were women, assuming I've done the maths right, you've got a 12.5% probability of the null hypothesis being right (two-tailed binomial distribution significance test). The 'usual' test for significance would be 5%, so we can conclude that these stats don't show a significantly increased risk for women.

Incidentally, if it were 7 out of 7, however, you're only on a 1.5% probability of the 'equally likely' hypothesis to be correct which would be significant at a 5% confidence level. So it's not necessarily the sample size which drives the assessment.

Look, I'm not a statistician, but it IS polishing a t.u.r.d.. Yes, there are statistical methods to address them, but it is better to have more data. Its a problem with stats, or any analytica method, to merely produce a sausage machine answer without considering the machinery to produce it. You can plug any numbers into a formula - the trick is to decide if its worth doing so.

rhext

Always Tyred wrote:

Look, I'm not a statistician, but it IS polishing a t.u.r.d.. Yes, there are statistical methods to address them, but it is better to have more data. Its a problem with stats, or any analytica method, to merely produce a sausage machine answer without considering the machinery to produce it. You can plug any numbers into a formula - the trick is to decide if its worth doing so.

It's always better to have more data. The question is how much data you need before you draw a conclusion. You make a statement which says a sample size of 7 is not enough, but you present no case as to how you might decide how big a sample size is required. How many times do you have to throw a die and get a 6 before you conclude that the die is likely to be loaded? The answer is actually 'it depends'. If you get all sixes, it's only about 4 or 5, but what if you get 1 six out of 5 throws. 10 sixes out of 50?. 100 sixes out of 500? The right sample size depends entirely on the level of bias of the dice. The thing is that 7 feels too small, but that's only true if men and women cycle roughly equivalent amounts. If (for example) men cycled 100 times as many miles as women and you had stats which showed 2 female deaths and one male, that sample may well be big enough to indicate that something strange was happening.

Why is this important? It's because you use these sort of stats to work out where to spend time and money on investigating what's going on. Is it worth a study to find out why women are disproportionately more likely to get sideswiped than men? On the basis of the two sets of stats presented here, it would seem to me very likely that there is a significant difference. Not sure how that's turd-polishing myself.

ellieb

Did someone say there are no statisticains on this board?

always_tyred

rhext wrote:

On the basis of the two sets of stats presented here, it would seem to me very likely that there is a significant difference. Not sure how that's junk-polishing myself.

If someone were now to post something suggesting that, in a much larger sampling (for example the whole UK over many years) men are proportionally more likely to be killed cycling, you might conclude that the 6/7 stat was in fact a statistically insignificant cluster, rather like flipping a coin and getting heads 6 times out of 7.

As such, you will only know if "very liklely" is correct when you have more information. For that reason, and its a very good reason, using limited sample sizes for a forward analysis is dangerous. At best, you might conclude that the issue was worth more study. Perhaps this is what you mean and we are talking at crossed purposes.

I have no idea if women are statistically more likely to have cycling accidents, but it would be interesting to know.

(Btw - I had to use "t.u.r.d." rather than "turd" because of the damn turd filter.)

benno68

http://www.guardian.co.uk/uk/2009/may/08/cycle-accidents-road-deaths-rise

Wonder if this reporter would investigate the issue we're discussing?

rhext

Always Tyred wrote:

(Btw - I had to use "t.u.r.d." rather than "junk" because of the damn junk filter.)

Yes, it looks like I got moderated, because I'm fairly sure I didn't type 'junk'..... :oops:

I'm not quite saying that. What I'm trying to do is answer the question 'if 6 incidents out of 7 involved females, how likely is it that is not down to blind chance'? If men and women do approximately equal mileage, you'd expect an equal number of incidents. The chances of getting 6 out of 7 in those circumstances is a little less than 1 in 10. On the other hand, if men do twice as many miles as women the chance of getting 6 out of 7 is much lower (I think it's less than 1 in 100). The point is that the maths allows you to determine the probability. After that, it's more about what your definition of 'very likely' is: 1 chance in 10? 1 chance in 100? 1 chance in 1,000?

moonio

I just had a thought after reading the very sad and distubing sticky thread about the Kennington accident RIP

Could it be that women somehow expect other road users to be courteours and considerate when they are cycling and vunerable..ie like men opening the door for them etc.
I must admit that when I started cycling I expected to be treated with a bit more delicacy than I have experienced.
Now I just imagine shooting the tyres off all HGVs I see :twisted: but in reality treating them with extreme caution

DonDaddyD

More men cycle that women.

More men are invovled in bike accidents than women.

When women are involved in bike accidents it resonantes louder because there are less women cyclists in general. - Think: a white guy won the 100meter sprint :shock:

I don't think women are more susceptible to bicycle related accidents than men. We are all equal, all vulnerable.

Cafewanda

Hi Moonio, I imagine shooting the tyres of every vehicle that "dun me wrong" :twisted:

I am a little concerned at how quickly I hurl verbal abuse in their direction though. My favourite word is 'Dick' :shock:

Jamey

I predict the final sentence of the above post will be quoted back at you forever more.

Kieran_Burns

Cafewanda wrote:

bla bla bla bla. My favourite word is 'Dick' :shock:

Quoting this so it can't be lost

DonDaddyD

Kieran_Burns wrote:

Cafewanda wrote:

My favourite word is 'Dick' :shock:

Quoting this so it can't be lost

Clearly a single speed mindset.... :shock:

tailwindhome

I point no fingers and don't exclude myself from this, but its a sobering thought that the guy who lost his wife is probably reading this thread.

This will make me thing twice about every post in the future.

Cafewanda

Jamey wrote:

I predict the final sentence of the above post will be quoted back at you forever more.

Oh hell. Didn't think of that :shock: :shock: :shock:

deptfordmarmoset

bump...

I just learned that someone has been charged in connection with the girl killed in Greenwich, Adrianna Skrzypiec, back in May. I'll try to find out more - all I know so far is what http://www.thegreenwichphantom.co.uk/ said in her/his blog today.