Talk:Bicycle helmets in New Zealand

Duynhoven's comments

The comments by Harry Duynhoven on helmet use in 2008 don't need to be documented yet but should be used for future commentary on the helmet debate. -- Alan Liefting (talk) - 22:54, 22 November 2008 (UTC)

Criticism and Support

Previously, there was only a Criticism section, the bulk of which was not New Zealand specific. Nonetheless, the criticisms do give the reader an idea of the tenor and gist of the debate. Therefore I left that material in but added some supporting arguments with references to help balance the section and the article as a whole. Nelsonsnavy (talk) 19:36, 2 February 2012 (UTC)

After further consideration, I removed any criticism and support arguments not specific to New Zealand, as they are covered in other helmet articles and could be misconstrued as somehow directly representing New Zealanders' views on the issue. Nelsonsnavy (talk) 13:45, 3 February 2012 (UTC)

The problem is that Nelsonsnavy deleted important information relating to NZ issues, e.g. that 3 NZ organisations (MOT, ACC, SafeKids) claim that helmets reduce brain injury by 88%. This figure is from a study published in 1989 that, to quote the main Wikipedia bicycle helmet page: "The most widely quoted case-control study, by Thompson, Rivara, and Thompson, reported an 85% reduction in the risk of head injury by using a helmet. There are many criticisms of this study.[48] Its figures of 85% and 88% protection from helmets are based on a simple misunderstanding of odds ratios, but are widely quoted. The positive findings have been attributed to confounding because the helmet wearers were a very different social group to the non-wearers. Cyclists on the streets at the time had about the same helmet-wearing rate as the cyclists with head injuries; cyclists with non-head injuries reported a higher rate of helmet use.[49]"

It seems an interesting and important aspect of the NZ helmet laws that that 3 main NZ road safety organisations don't quote NZ data on the benefits of helmets, but rely on an overseas study by Thompson, Rivara & Thompson (TRT) that made exaggerated claims based on a misunderstanding of odds ratios. Perhaps Nelsonsnavy can explain why this is not relevant, or why, on the Australian helmet law page, there is no mention in the quote included from the study commissioned by the Qld Gov, that it is a citation of TRT's claims, not Australian research. If the quote is included, readers need to know that it is a quote from TRT, rather than an evaluation of Australian helmet laws. (talk) —Preceding undated comment added 05:05, 5 February 2012 (UTC).

Still a work in progress. This article was filled with false info, argumentation, loaded language, incorrectly attributed sources, information with no bearing on New Zealand, outdated statistics, broken links, etc. Please see my individual edit comments. Please add info that is relevant, and I will try to do the same. But by reverting all the edits, you've reintroduced numerous problems into the article, severely compromising its objectivity, accuracy, and relevance. My goal is to provide factual data from reliable sources relating strictly to bicycle helmet law and use in New Zealand, not insert personal argumentation about why the law is good or bad.

Either way, the segment in question was not relevant. It was an editorial argument about a particular statistic. Such a passage would have bearing if, for example, a New Zealand advocacy group were claiming the government was using false statistics; then we'd need to quote that group's argument, not insert our own. Secondly, because one website--an anonymous commentary from an apparently partisan website, no less--has questioned a particular scientific study is neither here nor there. If there were broad scientific consensus that a particular study was flawed, that might be a different issue.

Finally, please note that the MOT stat you mention is sourced from a different study that the one originally disparaged in this article. SafeKids sources a different study, too, though by the same authors. ACC just says "American research." One therefore can't say they are all drawing from the same study that was mentioned before the edit. Nelsonsnavy (talk) 13:25, 6 February 2012 (UTC)

Reading through the edits, it's interesting that Nelsonsnavy keep removing the section relating to the origin of 85%/88% figure cited by these, perhaps the principal promoters of this law. And, yes, there is no other 'American research' that has ever given this figure; that is common knowledge among a decent slice of the global cycling community. Removing this section entirely seems akin to destroying unwanted evidence by an editor who, to all appearances, seems keen to remove all negative coverage of the helmet law under the guise of only including 'recent', 'NZ-relevent' references, ignoring its global context and the lack of awareness/debate on the issue in New Zealand. Is there a 'neutral' way to include this information? — Preceding unsigned comment added by 203.96.150.234 (talk) 09:30, 16 February 2012 (UTC)

Already explained above. It's a personal editorial argument. If NZ cycling groups, for example, are charging certain organizations with using false or misleading statistics, by all means include that here. I could not find any documentary evidence that any relevant NZ groups care about statistics proffered by Safekids et al. Further, if one document by a group quotes a statistic that has been criticized by one one web site, that is irrelevant; one can find a critic for anything. If a group were basing all their promotional efforts solely on one statistic that had been widely discounted by experts, that might well be "newsworthy," but even then, our goal is not to question them, but to report what they say. And after all, if you're going to inject your own point that someone has criticized a certain figure, then in fairness you need to quote arguments that support that figure.

I should add too that there is obviously no "destruction of evidence" (hyperbolic, don't you think?) because this is not a case or a trial. The law is in place already; some support it, some oppose it, some don't care. Our job as editors is merely to report that, not to make, bolster, or undermine arguments about the law or its proponents/opponents. What you or I think about the law is irrelevant. Nelsonsnavy (talk) 14:59, 16 February 2012 (UTC)

The fact remains that these groups cite that 'wearing a cycle helmet reduces the risk of head injury by 85%...'. This is a controversial statistic within the global cycling community. The uncritical use of this high figure (from 1987) without acknowledging that no other study has found such a high figure, that other research has found little or no benefit from helmets and that jurisdictions that have enforced mandatory helmet laws have seen little or no decline in head injury risk per cyclist (let alone an 85% reduction of risk) is highly questionable. As for being 'the law', laws change: it was NZ law to punish homosexuality until 1968; using a bicycle without a helmet is legal in most countries of the world. — Preceding unsigned comment added by 203.96.150.234 (talk) 00:36, 18 February 2012 (UTC)

I understand your point, but there are other things that need consideration: 1) this article is about cycle helmets in NZ only; the main bicycle helmet article deals with more general research findings and criticisms of them, including the figure you cite, and there are two articles here about helmet laws in general. This article is about not the global cycling community (most of whom probably know or care nothing about helmet laws or research, anyway) 2) who is questioning that statistic? Any cycling group or prominent activist or politician in NZ that has bearing on the topic at hand? 3) that study has also been reviewed and accepted by other scholars in their own work, either implicitly or tacitly accepting the findings--do we note that? (I'd say no because, as I said, I can't find evidence that it's an issue to involved parties) 4) you say, "other research has found little or no benefit from helmets," but OTOH there is lots of scientific evidence that helmets do provide protection (see the bicycle helmets article) 5) of course laws can change, but an encyclopedia is not even remotely the appropriate place in which to advocate for or against laws by trying to bolster or undermine the stances of involved parties. Nelsonsnavy (talk) 14:11, 18 February 2012 (UTC)

Anyone know of any New Zealand groups (cycling, medical, civil rights, etc.) who oppose helmet-wearing or the helmet law? Links appreciated. Thanks. Nelsonsnavy (talk) 17:31, 5 February 2012 (UTC)

http://www.stuff.co.nz/dominion-post/news/6395656/Helmet-law-halves-cyclist-numbers "Cycling Advocates' Network(CAN) spokesman Patrick Morgan who said helmet compulsion was a failed experiment... It does more harm than good and our policy is that it needs an independent review." Their policy page is at http://can.org.nz/helmets. Richard Keatinge (talk) 18:31, 12 February 2012 (UTC)

Thanks. I already found and included that in the article. They actually don't oppose the law, but rather call for it to be re-evaluated, though it's a low priority for the group and its membership at this time.Nelsonsnavy (talk) 12:45, 15 February 2012 (UTC)

Re-Write

Have finished a rough incremental re-write of this article to improve its objectivity, relevance, and timeliness. My goal was to keep the focus strictly on the relevant facts so someone researching the issue can get an unbiased overview. The article still needs polishing (particularly with regard to citation format uniformity) and one dead-link fix.

Any more NZ-specific studies, quotes, etc. most welcome. Article still needs info about the passage of the law itself: debate leading to it, political or economic effects thereafter, etc. Nelsonsnavy (talk) 14:52, 6 February 2012 (UTC)

Safekids NZ

I temporarily undid the edit about Safekids New Zealand's funding for multiple reasons: In the context, it seems to imply bias based on financial interest, something one should avoid imputing without facts--sort of liking saying in 1950's America, "Incidentally, one of their sponsors is a Communist." :) The reference is to a 4-year-old annual report for Safekids Worldwide. It is not clear if/how sponsor donations to the worldwide Safekids group make their way to the New Zealand organization ("a service of Starship Children's Health"). Seems like irrelevant info in this context, anyway: the point is to let the reader know what groups in New Zealand are taking stands about helmets, not who donates to them. This bears group discussion. Nelsonsnavy (talk) 19:02, 11 February 2012 (UTC)

OK. I've put it back. The reference makes it clear that the international organization claims credit for Safekids New Zealand, though they don't give exact details of their bank accounts and I don't think such details are required. The financial relationship is made quite clear - "$100,000 to $999,000" from Bell Sports - and is clearly nothing that they feel ashamed of. Nor should they - I can't for the life of me see that they are doing anything wrong or even remotely untoward, they are promoting child safety. I would suppose that their commercial sponsors are proud of their products and consider them effective. But the relationship is a salient fact and readers may draw what imputations they like from it. Richard Keatinge (talk) 19:10, 11 February 2012 (UTC)

I think we're talking at cross purposes: I was not implying (nor do I suspect) anything wrong or untoward on their part. Charities accept corporate donations all the time. My point is that adding that detail into the article seems, in the context of a hotly debated topic such as this, merely like an invitation for readers to cast suspicion on the New Zealand group's motivations for recommending helmets. And besides, it seems gratuitous since we don't make special note of the financial sources of the other groups discussed in the section. I'd like to find a compromise on this.Nelsonsnavy (talk) 19:25, 11 February 2012 (UTC)

May I suggest the compromise be that the reference stays but the wording be neutralised? Like it or not, that amount of money from a sponsor may indeed influence the recipient's behaviour and is therefore a salient fact. — Preceding unsigned comment added by 203.96.150.234 (talk) 19:38, 11 February 2012 (UTC)

As you say, "may" influence. Do we know it factually to be the case? I'm curious, too, why my edit of this is described as "apparent vandalism." It pointed to the Talk section so there could be discussion and consensus before casting doubt on a group's motivations. And considering I spent quite a few hours researching and fixing this article, I'm hardly going to vandalize it :) Nelsonsnavy (talk) 19:52, 11 February 2012 (UTC)

We don't know that this influences the stance of the organization so we're not saying that it does. It's still a salient fact. I'm not aware of any analogous salient facts in relation to any other organizations referenced here, but if there were any I'd definitely suggest mentioning them. And no, I don't think you've been vandalizing anything. In fact I'd like to thank you for your contribution to improving this article. Would you feel inclined to make a bold edit and return the sailent fact in words of your choosing? Richard Keatinge (talk) 21:57, 11 February 2012 (UTC)

I came up with neutral language that accords with what we know for sure of the facts, something like, "Safekids New Zealand, a national child safety organization, whose parent group has received donations from a helmet manufacturer, ..." Good compromise? Yes, I will be making a bold re-edit soon. Another editor has been systematically re-introducing verbiage that had been removed because it was editorializing and/or did not comply with Wiki guidelines for neutrality, verifiability, etc. Nelsonsnavy (talk) 13:38, 12 February 2012 (UTC)

Seems like a good compromise to me. Richard Keatinge (talk) 18:13, 12 February 2012 (UTC)

Happy with Richard's compromise words, except than they haven't yet appeared. I've also reinstated the fact that Safekids, the ACC and NZ MOT all claim that helmets reduce brain injury by 88%. Even though the attribution is sometimes vague "American research", or indirect (Macpherson and Spinks citing Thompson et al.) or a direct cite of Thompson et al., the source is still Thompson et al. It is relevant that NZ agencies chose to cite this study, rather than estimates from NZ. — Preceding unsigned comment added by Dorre (talk • contribs) 09:52, 25 February 2012 (UTC)

I stand by my contention that it's an extraneous editorial argument designed to undermine or call into question the position or data of certain groups when there's no evidence that any relevant New Zealand groups care about this particular issue. Further, it's solely working from arguments proffered by one advocacy site instead of using a balanced variety of neutral (or at least more neutral) sources. We need some sort of compromise: one would be to remove that paragraph and any references to Safekids et al. entirely. Another would be to note some of the scientific studies and medical experts that accept/incorporate the 88% figure, of which there are many. My vote is for the former course of action; otherwise the article will devolve further into a he-said-she-said polemic instead of merely presenting the relevant facts.Nelsonsnavy (talk) 14:18, 25 February 2012 (UTC)

Can we make progress towards a consensus by suggesting links to (rather than repeating arguments from) the appropriate sections of Bicycle helmet? In that case some brief comment might suffice here, perhaps something like "these organizations all base their arguments on an 88% figure, not from New Zealand, which has been questioned, see ..." Richard Keatinge (talk) 19:51, 25 February 2012 (UTC)

I imagine that's the best common ground we're going to reach on this one: just a simple statement with a reference.Nelsonsnavy (talk) 14:58, 26 February 2012 (UTC)

"Reintroducing the Australian connection."

Need a source for this. Page linked does not say how they're connected. It's just a list of countries and provinces.Nelsonsnavy (talk) 20:25, 11 February 2012 (UTC)

Misrepresentation of the results of Wang et al.

Linda Ward added the comment: “A 2013 study by Wang et al. identified a number of flaws in Robinson's re-analysis, showed that Povey's analysis had in fact taken the trends into account, and concluded that Povey's results were valid. [14]”

However, 1) Povey did not take time trends into account. 2) Wang’s model 4 (which Povey did not fit) accounted for time trends and showed no significant additional effect of helmets – the small non-significant estimate for helmets is positive, implying that if anything, increased helmet wearing increases head injury rates! Wang et al. also state “After checking the residual plots and test for serial correlation, it seems that the model assumptions for Model (4) are satisfied”

Wang’s model 4 is very similar to the models fitted in Robinson’s re-analysis, so rather than identifying flaws, Wang’s results appear to confirm the validity of Robinson’s finding that after accounting for time tends, there is no estimated benefit of helmets. Most Wiki readers will find it hard to understand Wang’s additional claim that both model 4 (which finds no benefit of helmets) and model 3 (Povey’s model, which ignores trends and therefore finds a benefit of helmets) can both be valid, especially since they also say that "a model with a linear time trend outperformed the one with helmet wearing rate as the explanatory variable.”

Wang’s describes Model 3 as Povey’s model, and compares it with models 4 and 5, which Wang describe as fitting time trends and outperforming model 3. Linda’s comment is obviously incorrect and misrepresents the results of Wang et al.

In the interest of accuracy and simplicity, I suggest replacing Linda’s obviously incorrect comment with something simpler: "A follow-up paper by Wang et al. confirmed that a model with a linear time trend outperformed the one with helmet wearing rate as the explanatory variable"

Wiki readers could tie themselves into knots trying to debate how two models that come to vastly different conclusions can both be “valid”, but we don’t want to inflict that sort of confusion on Wiki readers. The clear, simple, message from the Wang paper is that the model with the trend outperformed the model fitting helmets. Anyone disagree with my suggested redraft? Dorre (talk) 00:28, 21 December 2013 (UTC)

So, Wang actually found that long-term time trends accounted for changes better than helmet use. That seems clear enough, and we need to remove the weasel words. I agree with the redraft or with a shorter and perhaps clearer version, such as "Wang found that long-term time trends accounted for changes better than helmet use." Richard Keatinge (talk) 10:38, 24 December 2013 (UTC)

It might be even better to avoid the entire wrangle, based as it is on primary literature, and simply remove the sentence concerned. Richard Keatinge (talk) 17:40, 24 December 2013 (UTC)

Indeed. The entire section is the usual bollocks from people with conflicts of interest. One day, I'll summon the energy to get the lot of you topic banned. - Sitush (talk) 16:20, 27 December 2013 (UTC)

Would this be support for removing the sentence concerned? If so, let's do it. Richard Keatinge (talk) 16:31, 27 December 2013 (UTC)

I find myself roused from my Christmas revelry by a mysterious Wikipedia message (the sender forgot to log on so its from an IP address) requesting I come and look into this thread. Really you should all be off enjoying Christmas, but this latest issue is obviously weighing some of you down and probably detracting from your enjoyment of the season. Therefore in between the turkey and roast chestnuts I grabbed a few moments to download Wang and give it a quick scan... Here is a capsule summary, bear in mind that (a) I'm not here to agree/disagree with Wang - I leave that to the domain experts here if they so choose - just simply to try to report what it says, and (b) as I said its Christmas and my brain might be somewhat more addled that usual :-) :

The paper is a look at work related to New Zealand. Wang starts with an uncommon, to me at least, argument - while most, whatever their POV, seem to be of the opinion that pre-compulsion promotion, and compulsion, of bicycle helmet wearing produces a non-linear relationship between time and helmet wearing rates i.e. the promotion/compulsion boosts helmet wearing rates over and above what might be expected without them, Wang argues there is in fact a linear relationship between the two - i.e. they appear to argue that promotion/compulsion fails in its goal of increasing wearing rates.

Wang then, having also argued that time trends need to be taken into account - one of Robinson's points, gets into a rather tortuous section in which they argue that Povey, who didn't take into account time trends, did in effect take account of them as they took account of helmet wearing rates which are linearly related to time... So Robinson was right, as Povey didn't knowingly take account of time trends, but as Povey unknowingly and indirectly did then Robinson is also wrong and Povey's results are indeed valid. However this is not Wang's final statement on this issue, immediately after they state that time trends alone actually explain the changes in injury rates better, i.e. Robinson is in fact right, or at least a better fit than Povey... Or something like that, as I said it all gets rather tortuous and hard to disentangle.

Wang follows the comments on Povey & Robinson with looks at other work related to New Zealand by Scuffam, Langley and Clarke - these are less detailed, in a statistical sense, then the first section and there is a certain opinion/commentary flavour to these.

In their conclusion Wang et al quickly move from New Zealand to the Atwell/Elvik/Church series and discuss that - it is almost as though we're reading a conclusion from another paper altogether.

So back to the Wikipedia article content.

First the reference to Wang is just a link to a PDF, no bibliographic data at all - that needs fixing.

Second, Ward's summary is clearly incorrect, referring as it does to the "valid" without the immediately following "however". But to be fair attempting to summarise Wang is I suspect a challenge for anyone (which might explain why their own conclusion seems to go off on a tack). Therefore what has been added needs to go as it is incorrect, and what, if anything is added in its place can be discussed by the experts here on Talk first.

Thirdly, the Wang paper is referenced twice. The second reference refers to the publishers of the paper, rather than the authors, and states they "noted" something about Clarke. Well they discussed/argued/present an opinion on Clarke, they didn't "note". That needs a little tidy up as well.

I'll go make those edits and then return to Christmas revelry. I suggest you all do the same, leaving discussion on replacement text till the New Year.

So Happy Christmas to all of you. I look forward to dropping by later in 2014 and seeing you all here, whatever POV camp you are in, constructively discussing amongst yourselves and building a better article! Kiwikiped (talk) 02:48, 28 December 2013 (UTC)

Yes, The Wang paper is rather convoluted and difficult to follow. This quote from the paper seems to summarize what they are trying to say: “Hence, even the hypothetical data contains no effect of helmet wearing by construction, the variable HELMET is highly significant in predicting the ratio of head to limb injuries because both the variable and the outcome can be predicted using a linear function of TIME.”

It’s as if they have fallen into the trap that correlation implies causation. Just because helmet wearing increases with time, and the hypothetical dataset was constructed with a downward time trend, doesn’t imply that one caused the other. Yet most people would expect a claim that the model is “valid” to imply there is evidence of causation!

Wang et al. also shoot themselves in the foot in the last paragraph of that section by saying: “it seems that the model assumptions for Model (4) are satisfied and hence the results and conclusions in Povey et al.’s analysis are valid.” I believe this is a careless typo and they meant to say Model (3), the model Povey actually fitted. It doesn’t say very much for the paper (or the peer-review process) that such a key issue that affects the meaning of the entire paper wasn’t corrected. I'll leave it to other editors to decide whether it would be helpful or informative to cite this paper. Dorre (talk) 07:44, 5 January 2014 (UTC)

In response to various comments above by Doree, Kiwikiped and Richard Keatinge . . .

Doree's claim that "Povey did not take time trends into account" is incorrect, and her claim that Wang's results "appear to confirm the validity of Robinson's finding that after accounting for time trends, there is no estimated benefit of helmets" GROSSLY misrepresents Wang's conclusion that "the results and conclusions in Povey et al.’s analysis are valid".

It seems that Doree does not understand that, as clearly documented by Povey et al., time trends were (implicitly) accounted for by including limb injuries in the models. A manual calculation clearly shows that Povey's approach DID take time-trends into account. In 1994-96, compared to 1990-93, limb injuries ROSE by 12.6%, and head injuries FELL by 16.6%. This equates to a 29% drop in head injuries, after the 'time-trend' has been taken into account. (This is 'amazingly' close to both the 28% estimated by Povey, and Walter's results. Walter et al. explicitly controlled for the time trend by including a TIME term in their models, which did not cause a collinearity problem because they used a law indicator variable instead of a helmet wearing percent. If Povey's model had NOT taken the increase in limb injuries into account, Povey would have estimated that head injuries fell by only about 17%.)

Kiwikiped's claim that "Robinson was right, as Povey didn't knowingly take account of time trends" is clearly incorrect, if Kiwikiped were to peruse Povey's paper he would see that Povey DID knowingly take time trends into account: "Some measure of cyclist exposure to the possibility of head injury was required, to take into account any variations in the level of risk over time . . . It reflects both the amount of travel done by cyclists and changes to the cycling environment, which include changes in the nature of cycle use and safety improvements in the overall road environment.".

Doree's claim/s that "Most Wiki readers will find it hard to understand Wang’s additional claim that both model 4 (which finds no benefit of helmets) and model 3 (Povey’s model, which ignores trends and therefore finds a benefit of helmets) can both be valid" misrepresents both the Povey and Wang papers. Wang et al. did NOT claim that model 4 was valid, they said that the model ASSUMPTIONS (of no serial correlation) were SATISFIED, and that "the RESULTS and CONCLUSIONS in POVEY et al.’s ANALYSIS are VALID". They also noted that "the variable HELMET is highly correlated with time (r=0.90)", and explained why including both (such highly correlated) variables in the model would cause a multi-collinearity problem. Doree obviously does not understand collinearity, gotten herself confused over model 4, and incorrectly interpreted the helmet (small positive) coefficient and (non-significant) p-value as "implying if anything, increased helmet wearing increases head injury rates!". As Doree did not understand Wang et al.'s explanation of collinearity, the links below (or an introductory modelling course) may help her to untie the knots she has got herself into "trying to debate how two models that come to vastly different conclusions can both be 'valid'". (ie. understanding collinearity should help Doree to find the Wang paper less "difficult to follow"; understand why Povey' results and conclusions are valid; and why, whilst the ASSUMPTION of no serial correlation is valid, model 4 results are NOT, because of collinearity between TIME and HELMET).

• http://learnitdaily.com/six-ways-to-address-collinearity-in-regression-models/ - "Standard linear regression methods are known to fail (or, at least, perform sub-optimally) in the presence of highly correlated predictors"
• http://en.wiki.x.io/wiki/Multicollinearity - "a multiple regression model with correlated predictors can indicate how well the entire bundle of predictors predicts the outcome variable, but it may not give valid results about any individual predictor, or about which predictors are redundant with respect to others"
• http://homes.ori.org/~keiths/Files/Methods/Multicollinearity.html - reduced stability of the corresponding parameter estimates, increased standard errors, and reduced power to measure effects

Wang does not "argue" that there is in fact a linear relationship between the two, Wang NOTES the FACT that a (simple) linear regression of helmet wearing on time shows that there IS a linear relationship between the two. The linear regression estimates that, as shown in Fig 1b, between 1990 and 1996, helmet wearing increased on average by 12% per year (p=0.006, r2=0.81).

Wang et al. do NOT (appear to) argue that promotion/compulsion fails in its goal of increasing wearing rate, they note that "The helmet wearing rate increased from virtually zero in 1986 to 84%, 62% and 39% in September 1992 for primary school children (5-12 years of age), secondary school children (13-18 years) and adult commuters (over 18 years), respectively (Scuffham, et al., 2000). Just after the legislation, helmet wearing rates increased to above 90% for all cyclist age groups."

Fig 1b clearly shows that whilst (adult) helmet wearing was increasing by about 5% per year with promotion, it jumped from about 40% in 1993 to about 90% in 1994. Fig 1b also clearly shows that whilst the linear regression perfectly predicts the 1991 wearing rate, and predicts the 1990, 1992, 1995 and 1996 to within 10%, it is not a good fit for the 1993 and 1994 values. However, the r2 of 0.81 means that 81% of the variation in helmet wearing is explained by the linear relationship with time, with helmet wearing increasing on average 12% per year (p=0.006).

Wang did NOT, as claimed by Kiwikiped, echoed by Richard Keatinge and 'confirmed' by Doree, find "that long-term time trends accounted for changes better than helmet use", they found that "a model with a linear time trend outperformed the one with helmet wearing rate as the explanatory variable" and that "including both time trend and helmet wearing rate in the model may cause a multi-collinearity problem since they are highly correlated". Wang et al. reported an r2 of 0.69 for Povey's model and 0.90 for the model using (only) a linear time trend. This means that (about) 70% of the reduction in the head to limb ratio is explained by helmet wearing (reducing head injuries by 28%), and that 'time trends' were responsible for an additional 20%, hence "the results and conclusions in Povey et al.’s analysis are valid".

This part of Doree's response to Richard Keatinge is an egregious misrepresentation of Wang et al.: "It’s as if they have fallen into the trap that correlation implies causation. Just because helmet wearing increases with time, and the hypothetical dataset was constructed with a downward time trend, doesn’t imply that one caused the other. Yet most people would expect a claim that the model is 'valid' to imply there is evidence of causation!"

The point Wang et al. were making was the COLLINEARITY between HELMET and TIME, but Doree obviously does not understand that collinearity (a concept taught in introductory modelling courses) refers to CORRELATION between TWO PREDICTOR variables. In pointing out that Dorre's claim of a "spurious" association was obviously incorrect, Wang et al. in NO WAY suggested causation: "To demonstrate the potential bias associated with failure to fit time trends, Robinson created some hypothetical data containing no effect of helmet wearing. The data contains only a linear trend in which the ratio of head to limb injuries falls by 0.1 every year. Robinson then fits the following linear model for this data . . . and obtained a highly significant estimate for â , which Robinson claimed was 'spurious'. However, this result is not as “spurious” as Robinson claimed, since the variable HELMET is highly correlated with time (r=0.90)."

By "Yet most people would expect a claim that the model is 'valid' to imply there is evidence of causation!", Doree has not only demonstrated that she does not understand collinearity, she has also misrepresented the "valid" as applying to the model RESULTS, when in fact Wang clearly stated that it was the model ASSUMPTIONS that were valid. They also described why model 4 results are NOT valid: "The problem with fitting a model with both HELMET and TIME is multi‐collinearity: the predictor variables HELMET and TIME are highly correlated. We believe this is the case here because a multiple regression finds an insignificant estimated coefficient of HELMET and yet a simple linear regression on this variable shows the estimate is significantly different from zero. One of the consequences of multi-collinearity is that while controlling for other variables (for example TIME), the estimate of a variable (such as HELMET) tends to be less precise, hence its influence on the dependent variable cannot be as accurately estimated".

Wang et al. do NOT "shoot themselves in the foot", as claimed by Doree, their observation that "the model ASSUMPTIONS for Model (4) are satisfied and hence the results and conclusions in Povey et al.'s analysis are valid" is NOT "a careless typo" as claimed by Doree.

As noted in the Wikipedia (multicollinearity) link above "a multiple regression model with correlated predictors can indicate how well the entire bundle of predictors predicts the outcome variable, but it may not give valid results about any individual predictor, or about which predictors are redundant with respect to others ". If the assumptions for model 4 (containing the "redundant" time variable) are satisfied, then the assumptions for Povey's model are also satisfied. Doree claims that "It doesn't say very much for the paper (or the peer-review process) than such a key issue that affects the meaning of the entire paper wasn't corrected", however it actually Doree's lack of understanding of key (basic) statistical modelling concepts that has caused her to misunderstand/misrepresent the meaning of the paper.

In summary

• Wang et al. did NOT claim that model 4 was valid, they noted that whilst the model ASSUMPTIONS were valid, the model RESULTS were NOT, because of mulicollinearity
• the clear, simple conclusion from the Wang paper is that "the results and conclusions in Povey et al.’s analysis are valid"

Some further evidence of Doree's lack of understanding of statistical (loglinear time-series) modelling . . .

http://cyclehelmets.org/1241.html (by DL Robinson) incorrectly states that Carr et at. "didn’t bother to check that the same trend was evident for pedestrians, so had nothing to do with helmets!". In fact, Carr et al. clearly stated that the DID check, and adjust, for trends . . .

p3-4: "Firstly trends in head injuries were investigated . . . Hospital Admissions data were examined from overall trends (for all hospital admissions, not only road users), as well as those for bicyclists and pedestrians."

p4: general road safety countermeasures, speed cameras , random breath testing, impact on road safety, "To more closely ascertain the impact of these more or less coincidental counter measures, one can analyse data from two groups of road users that have much common ground with bicyclists: pedestrians and motorcyclists . . . In particular pedestrians are more likely to be the suitable comparison group. . . previous MUARC projects have used bicyclists without a head injury as a correcting factor for the effects of varying exposure . . . In this analysis numbers of persons without a head injury will also be incorporated into the model . . .".

p10:"In the modelling of monthly numbers of head injured bicyclists admitted to hospital the various predictors shown in Appendix 6.3 were entertained for inclusion in the explanatory model. These included numbers of non-head injured bicyclists admitted to hospital . . . numbers of head injured pedestrians . . .".

http://www.cyclehelmets.org/1228.html contains an obvious factual error with respect to the Walter study. As the only BHRF editorial board member with any statistical qualifications, one would expect that Doree would have been responsible for preventing the publication of the obviously incorrect statement that "Walter et al developed a statistical model and conclude a 25% to 29% reduction in head injuries to cyclists (relative to pedestrians) when the helmet legislation came into effect" - Walter et al.'s model compared cyclist head injuries to cyclist limb injuries, not pedestrian head injuries, the reductions reported were relative to cyclist limb injuries, not pedestrians.

The BHRF 'critique' also complains that "they do not explain how the figures were derived", and Doree made the same complaint in the "Doree's misinterpretation of statistical results" talk section that Jake Olivier created to correct Doree's misinterpretations of the Walter study. In the same discussion, Doree claimed that "Most reviewers of journal papers expect tables to make sense independently of the text. If I sent round a paper like that to my work colleagues, it would not pass the internal review and I would not be allowed to submit it for publication." Introductory log-linear modelling courses teach students how to derive such figures, as noted by Olivier "It's just simple algebra". (The paper passed the Accident Analysis and Prevention peer review, presumably because the AAP reviewer/s is/are familiar with log-linear modelling, which is the standard technique for modelling count data.)

Whilst the site if bursting with examples that don't "say very much" for the BHRF 'peer review' process, the claim that "Serious injury due to land transport accidents, Australia, 2003-04" (http://www.nisu.flinders.edu.au/pubs/reports/2007/injcat107.php) shows that "wearing a helmet seems to have no discernible impact on the risk of head injury" almost renders me speechless.

It also reminds me of a claim by BHRF Editorial Board Member Guy Chapman, that "50% of cyclist deaths in London are due to crushing by goods vehicles at junctions, cause of death being abdominal trauma" (http://en.wiki.x.io/wiki/Wikipedia:Reliable_sources/Noticeboard/Archive_146). I replied to Guy, asking him to provide a reference, but he never replied (sigh). I also cited http://www.rospa.com/roadsafety/adviceandinformation/cycling/facts-figures.aspx as showing that

• about 75% of cyclists killed have major head injuries
• chest and abdomen injuries occur much less frequently (5%), but are often serious. When they do occur they are often accompanied by head injuries.
• a study by the Transport Research Laboratory found that over 70% of the cyclist fatalities in London had moderate or serious head injuries, and over 80% of cyclists killed in collisions on rural roads had moderate or serious head injuries.

Which reminds me of an article, co-authored by another BHRF Editorial Board member (Wardlaw), http://www.ncbi.nlm.nih.gov/pubmed/23227191, which was recently cited by Doree (http://theconversation.com/politics-trumps-hard-headed-reason-on-bicycle-helmets-20973#comment_272400). This observer http://www.anenglishmanscastle.com/archives/cat_health_and_safety_6.html (rather aptly) described the article as "bollocks".

Another apparent misrepresentation I recenty noticed is the claim that a 2012 paper by Fyhri showed that "interventions such as helmet laws have no net benefit", which appears to be the BHRF's 'interpretation' of Fyrhi's observation that "reduced cycling will quite clearly have negative social health consequences".

Just because the BHRF site is bursting with misrepresentations does not mean that the Wikipedia helmet articles should reflect the same (anti-helmet) bias.

Linda.m.ward (talk) 11:52, 2 March 2014 (UTC)

In the above posting Linda.m.ward refers to my earlier writing:

"So Robinson was right, as Povey didn't knowingly take account of time trends, but as Povey unknowingly and indirectly did then Robinson is also wrong and Povey's results are indeed valid."

writing:

"Kiwikiped's claim that "Robinson was right, as Povey didn't knowingly take account of time trends" is clearly incorrect, if Kiwikiped were to peruse Povey's paper he would see that Povey DID knowingly take time trends into account"

Stirred from my Christmas revelry I reviewed Wang, I am happy to acknowledge I did not read every reference Wang makes to determine if their representation of that reference was actually correct.

Wang reports that Robinson states that Povey did not take into account time trends. Wang then did not simply say "Robinson has misrepresented Povey", as Ward is now doing, they went to some effort to argue that Robinson's argument was invalid because Povey indirectly took account of time trends.

So if Ward is correct, and I have not checked, that Povey directly and knowingly took time trends into account then Wang apparently misunderstood Povey and went to some effort to argue something they had no need to as it was written plainly in Povey's text. In other words Ward is now indirectly arguing that Wang's, not my, representation of Povey is wrong...

I am left wondering why anybody, regardless of POV, would wish to reference Wang.

Kiwikiped (talk) 21:15, 2 March 2014 (UTC)

Kiwikiped might well wonder why anyone would wish to reference Wang! Some insight into the wider picture can be gleaned by looking at http://www.cycle-helmets.com/helmet-law-spin.html in particular the graph in section 4) showing the remarkable similarity between the time trends for pedestrian and cyclist head injuries. It seems highly likely that the two have a common cause and that any attempt to fit a model explaining the trend in cyclist injuries by increased helmet wearing would not represent the obvious reality of the situation.

The same applies to Povey's data. Povey did not take the sort of time trends illustrated in http://www.cycle-helmets.com/imgs/spin-9.gif into account, so if the trends in NZ had a similar cause to those in Victoria, in the normal sense of the word, Povey's analysis would be considered invalid. Wang seems to be arguing that if A is correlated to B, then a model of B that fits A is valid. Irrespective of whether or not this is true, "validity" in this sense cannot be interpreted as causation. Indeed, Wang's analysis of Povey's data shows that, although we can't be certain which is the cause, the a simple linear trend is much more likely.

In summary, Wang's definition of validity should not be taken as evidence of causation, any more than correlations imply causation. By presenting these results in a way that might lead people might think there was evidence of causation, Linda misrepresented the results of Wang et al.Dorre (talk) 11:36, 3 March 2014 (UTC)

We certainly don't need to inflict this kind of thing on readers of an encyclopedia. We do, however, need to quote Wang because of her specific comment on Clarke. Richard Keatinge (talk) 12:19, 3 March 2014 (UTC)

Removal/replacement of material inconsistent with the intention of the source

http://en.wiki.x.io/wiki/Wikipedia:Synthesis#Synthesis_of_published_material_that_advances_a_position advises: Take care not to go beyond what is expressed in the sources, or to use them in ways inconsistent with the intention of the source

This is particularly relevant with respect to the citation of the analysis by Elvik, which included data on 47 fatalities, mainly in the US. None were in NZ. Results for fatalities are contained in one line of one table in the paper. Elvik does not attempt to draw any conclusions from this small sample. In fact, Elvik is quoted discussing his research at http://sciencenordic.com/bike-helmets-are-less-effective-we-think This web page clearly shows that the use of this one line of a table buried in the middle of the paper is inconsistent with the views of the author of the paper. I will therefore delete this line, and replace it with a quote on the effects of helmet wearing from a much more relevant review into a larger number (84 cyclists) killed on roads and pathways in NZ in the years 2006-12. I will also abbreviate the quote from a paragraph buried in the middle of the Scuffham paper and extend the quote from the abstract, which I think more closely represents the intention of the author. Dorre (talk) —Preceding undated comment added 07:52, 5 January 2014 (UTC)

Doree claims that the 6 case-control studies in the Elvik meta-analysis was only a "small' sample, when the 95% CIs showed a statistically significant effect, yet she has replaced it with a 'review' that is GROSSLY under-powered (ie. a MUCH larger sample size would be required to enable any inferences to be made).

Koorey's "much more relevant review" has not been peer-reviewed. In the 1st reference cited by Doree, http://can.org.nz/system/files/CoronerInquest-Notes-GKoorey-v2.4-Jun2013.pdf, Koorey describes his review as "notes"; the 2nd reference is merely an abstract for a POSTER presentation at a conference.

If Koorey had calculated a (95%) confidence interval (CI), he would have seen that the upper bound was 18%, which means that the 'true' non-wearing rate for in fatal crashes could be as high as 18% (ie. more than double the national non-wearing rate). Koorey apparently has a PhD in engineering, but no statistical qualifications http://www.development.org.nz/about-us/our-presenters-coaches-and-consultants/personInfo?person=47, so it is not surprising that he made such a fundamental error. However, given that Doree is a "senior statistician" (http://www.ncbi.nlm.nih.gov/pmc/articles/PMC1410838/pdf/bmj3320722a.pdf), it is 'inexplicable' that she failed to notice Koorey's (grossly) inadequate sample size (only 9, yes NINE, unhelmeted cyclists?!).

A recent study by Persaud et al. (http://www.cmaj.ca/content/184/17/E921.long), which was not 'biased' by a grossly inadequate sample size, found that unhelmeted cyclists had more than 3 times the risk of dying from a head injury than helmeted cyclists (OR=3.1, 95% CI 1.3-7.3, for those with only a head injury and no other substantial injuries the OR was 3.6, the 95% CI 1.2-10.2)

Another recent study, by Gomei et al. (http://www.ncbi.nlm.nih.gov/pubmed/23945263), concluded that "To effectively reduce bicyclist fatalities from traffic accidents, helmet use should be required for all bicyclists".

In his "much more relevant review" Dr Koorey fails to provide any evidence to support his claim that "helmets are generally no protection to the serious forces involved in a major motor vehicle crash". Public health research courses teach students to cite their evidence, and that a literature review is a critical part of the research process. If Dr Koorey had done a literature review, he should have encountered the Bambach study (http://www.ncbi.nlm.nih.gov/pubmed/23377086), which found that "Helmet use was associated with reduced risk of head injury in bicycle collisions with motor vehicles of up to 74%, and the more severe the injury considered, the greater the reduction."

It is unethical for (human) research to be conducted by individuals who lack appropriate qualifications (http://www.nhmrc.gov.au/_files_nhmrc/publications/attachments/r39.pdf), it is also unethical for research to be conducted using an inadequate sample size (http://www.nhmrc.gov.au/_files_nhmrc/publications/attachments/e72_national_statement_131211.pdf).

Re Doree's claims that citing Elvik's results are "inconsistent with the intention of the source" . . .

According to Elvik's abstract, the purpose of his analysis was to determine the extent to which Attewell's study was "influenced by publication bias and time-trend bias that was not controlled for". Elvik noted that "adjusting for publication bias has a very small effect on summary estimates of effect. Attewell et al. (2001) were therefore correct in their conjecture that the possible presence of publication bias did not greatly influence summary estimates of effect".

In addition to showing the "very small effect" that publication bias had on Attewell's results, Elvik's table 3 shows that all the fatal injury results are statistically significant, and that (in the event of a crash), compared to helmeted cyclists, unhelmeted cyclists were

• 2.3 times more likely to sustain a head injury
• 2.5 times more likely to sustain a brain injury
• 4.3 times more likely to sustain a fatal (head) injury

In claiming that http://sciencenordic.com/bike-helmets-are-less-effective-we-think shows that citing Elvik's fatality (and brain injury result) is "inconsistent with the views of the author of the paper", Doree seems to have overlooked the fact that Elvik is quoted as saying that "if we look at the individual injuries . . . there are indications that the helmets provide better protection against serious injuries than against less serious ones".

The fact that Elivk's results are not inconsistent with BHRF ideology is not a valid reason for suppressing the important information re BIG reductions in the risk of brain and fatal (head) injuries.

Most cyclist fatalities caused by head injuries, the Tin Tin data shows that in 1996-99 (after NZ helmet law) compared to 1988-91 (before law), per hour cycled, traumatic brain injury (TBI) rates DECREASED, when upper extremity injury (UEI) rates INCREASED

• for AIS>=3 injuries with motor vehicle (MVI) involvement, the TBI admission rate decreased by 63%, the UEI admission rate increased by 50%
• for AIS>=3 injuries with no MVI involvement, there was no change in the UEI admission rate, the TBI admission rate decreased by 76%
• for AIS<3 injuries with MVI involvement, the TBI admission rate decreased by 31%, the UEI rate admission rate increased by 100%
• for AIS<3 injuries with no MVI involvement, the TBI admission rate decreased by 38%, the UEI admission rate increased by 46%

A participant at http://theconversation.com/politics-trumps-hard-headed-reason-on-bicycle-helmets-20973 recently commented that "Mr Clarke, I have suffered a head injury, although not through cycling, that has left me sufficiently debilitated as to irreparably affect my life and career. If I had suffered that injury while cycling without a helmet, and information about the increased risk of not wearing a helmet had not been made available to me, then I would consider the withholding of that information unconscionable. This conclusion frames all further communication between you and I, and between myself and Dr Dorothy Robinson."( http://theconversation.com/politics-trumps-hard-headed-reason-on-bicycle-helmets-20973#comment_283476).

It seems unlikely that the (unconscionable) withholding of Elvik's brain and fatal (head) injury results would be consistent with Wikiedia policy.

Linda.m.ward (talk) 11:28, 2 March 2014 (UTC)

The fact that correlation does not imply causation may be a tired and old cliché, but it’s still just a relevant as ever!

When helmet wearing rates were low, the initial adopters were often children learning to ride and also mountainbikers. Falling off a bike in those circumstances (off-road, or in areas where kids learn to ride, usually where there are few, if any motor vehicles) is hardly likely to carry the same risk of a fatal injury as cycling on busy roads.

Elvik’s re-analysis found significant time trend bias – new studies (that generally involved a higher proportion of cyclists wearing helmets and so were more representative of normal cycling) found the benefits of helmets were much lower than previously thought. Although today’s helmets might be much less effective than older models, it seems more likely that the early studies suffered from confounding in that the cyclists who chose to wear helmets were in different types of accidents to cyclists who did not wear them. This was a recognised problem for other studies (e.g. hormone replacement therapy) that used similar methods.

In New South Wales, from 1992-94, where helmet wearing was known, 80% of fatally injured cyclists were wearing helmets. Helmet wearing in the NSW surveys at this time found that 84% for adults and 75% of children wore helmets. This (like the data on post-law fatalities in NZ) strongly suggests that helmet wearers had similar risk of fatal injuries to non-wearers.

Elvik wasn’t able to provide any adjustments for time trend bias in fatality data, because he didn’t have any new studies. I’ll write and ask if he supports the context in which Linda has used his results, or whether he thinks she is misrepresenting the truth by citing results that Elvik’s work suggests might be biased because of time trends.

I am concerned that Linda’s POV-pushing might also break the Wiki rules on synthesis. The abstract of the Taylor & Scuffham paper concludes that “large costs were imposed on adult (>19 years) cyclists”. This isn’t consistent with Linda’s POV, so she trawls through the discussion to find a sentence speculating that the costs might have been under-estimated: “Consequently, the minimum value of preventing a permanently disabling head injury is $2 million—the value of preventing one fatality”. This appears to be speculation - there is no evidence in the paper that any permanently disabling head injuries were actually prevented. The authors also caution “However, bicycle helmets do have some limitations. For example, the effectiveness of helmets is reduced where collision forces are greater than 30 km per hour.”

Linda takes an extract from Elvik’s paper (that didn’t include any NZ studies, and which Elvik would probably argue was biased because the estimate could not be adjusted for time trends) and juxtaposes it against a sentence in the discussion of the Taylor & Scuffham paper in a manner that contradicts the main conclusion in the abstract of that paper (large costs were imposed on adult (>19 years) cyclists). Do people think this is a case of synthesis and POV-pushing? Dorre (talk) 11:56, 3 March 2014 (UTC)

A simple solution might be to omit the Elvik paper entirely; as you say it doesn't include NZ data and doesn't specifically illuminate the NZ debate. Richard Keatinge (talk) 12:19, 3 March 2014 (UTC)

I have removed it and made a couple of other changes. Richard Keatinge (talk) 14:21, 3 March 2014 (UTC)

External links modified

Hello fellow Wikipedians,

I have just modified 3 external links on Bicycle helmets in New Zealand. Please take a moment to review my edit. If you have any questions, or need the bot to ignore the links, or the page altogether, please visit this simple FaQ for additional information. I made the following changes:

• Added {{dead link}} tag to http://www.transport.govt.nz/research/Cyclehelmetuse2011/
• Corrected formatting/usage for http://www.safekids.org.nz/Downloads/Safekids
• Corrected formatting/usage for http://www.transport.govt.nz/research/Documents/Cyclist-crash-fact-sheet-2010.pdf
• Added archive https://web.archive.org/web/20081014162428/http://bikenz.org.nz/Article.aspx?ID=1037 to http://www.bikenz.org.nz/Article.aspx?Id=1037

When you have finished reviewing my changes, please set the checked parameter below to true or failed to let others know (documentation at {{Sourcecheck}}).

This message was posted before February 2018. After February 2018, "External links modified" talk page sections are no longer generated or monitored by InternetArchiveBot. No special action is required regarding these talk page notices, other than regular verification using the archive tool instructions below. Editors have permission to delete these "External links modified" talk page sections if they want to de-clutter talk pages, but see the RfC before doing mass systematic removals. This message is updated dynamically through the template {{source check}} (last update: 5 June 2024).

• If you have discovered URLs which were erroneously considered dead by the bot, you can report them with this tool.
• If you found an error with any archives or the URLs themselves, you can fix them with this tool.

Cheers.—InternetArchiveBot (Report bug) 20:49, 1 November 2016 (UTC)

External links modified

Hello fellow Wikipedians,

I have just modified 3 external links on Bicycle helmets in New Zealand. Please take a moment to review my edit. If you have any questions, or need the bot to ignore the links, or the page altogether, please visit this simple FaQ for additional information. I made the following changes:

• Added archive https://web.archive.org/web/20140603194234/http://journal.nzma.org.nz:80/journal/abstract.php?id=5046 to http://journal.nzma.org.nz/journal/abstract.php?id=5046
• Added archive https://web.archive.org/web/20130503205229/http://archive.can.org.nz/ecan/ecan61.htm to http://archive.can.org.nz/ecan/ecan61.htm
• Added archive https://web.archive.org/web/20110724201036/http://spokes.org.nz/article/minister-for-transport-safetys-comments-on-revising-the-bicycle-helmet-law to http://spokes.org.nz/article/minister-for-transport-safetys-comments-on-revising-the-bicycle-helmet-law

When you have finished reviewing my changes, you may follow the instructions on the template below to fix any issues with the URLs.

This message was posted before February 2018. After February 2018, "External links modified" talk page sections are no longer generated or monitored by InternetArchiveBot. No special action is required regarding these talk page notices, other than regular verification using the archive tool instructions below. Editors have permission to delete these "External links modified" talk page sections if they want to de-clutter talk pages, but see the RfC before doing mass systematic removals. This message is updated dynamically through the template {{source check}} (last update: 5 June 2024).

• If you have discovered URLs which were erroneously considered dead by the bot, you can report them with this tool.
• If you found an error with any archives or the URLs themselves, you can fix them with this tool.

Cheers.—InternetArchiveBot (Report bug) 22:19, 13 January 2017 (UTC)