I love helmets

[bianchi1]

This shows up the flaws of case control studies rather well, as @srw has already mentioned. The numbers are very small. If 18% of the cyclists in the study wore helmets, you'd expect to see about 35 of them with head injuries. The reported number was 34, well within the expected standard deviation (a measure of how the numbers will vary through randomness). In other words, the raw data does not support the idea that helmets reduce head injuries. In fact, with so few data, I'd be surprised if any of the variations in this graph were statistically significant. In other words, any differences are probably due to random variation.

Worse still, having had a quick scan through the article, the figures I'm looking for (Chi-squared number, something which indicates whether the observed difference is due to chance or may actually be real) are missing. Why not - they used chi-squared analysis elsewhere? This is a very poor paper. It does not establish any benefits to helmet wearing. Quite the opposite - injury rates for both groups are rather similar: there is nothing statistically significant. Had it passed across my desk for review, I'd have rejected it.

I was always suspicious of conference papers...often an agenda. I guess this is why it's freely available on line, rather that a 30 quid fee to read.

 

[srw]

Worse still, having had a quick scan through the article, the figures I'm looking for (Chi-squared number, something which indicates whether the observed difference is due to chance or may actually be real) are missing. Why not - they used chi-squared analysis elsewhere? This is a very poor paper. It does not establish any benefits to helmet wearing. Quite the opposite - injury rates for both groups are rather similar: there is nothing statistically significant. Had it passed across my desk for review, I'd have rejected it.

Which in itself is an issue. If only "interesting" answers are published, a skewed view of the world results. Statistically speaking, one in 20 tests will give a <5% p-value. But all 20 need to be published so that the 1-in-20 doesn't become the truth.

 

[threebikesmcginty]

Most academics are rather keen to get their work published - before someone else does! It typically takes a few months to get through the review process, so the published articles aren't really out of date. Though I do seem to be straying off the point here somewhat...

I may be able to get access to the article - I'll look for it the next time I'm working at Liverpool University (some time this week).

Yeah right!

 

[Wobblers]

Which in itself is an issue. If only "interesting" answers are published, a skewed view of the world results. Statistically speaking, one in 20 tests will give a <5% p-value. But all 20 need to be published so that the 1-in-20 doesn't become the truth.

I wouldn't reject this paper for not showing any statistically significant results - I'd reject it for not giving the p-values. Actually, they do give p-values for one injury class - facial injuries, p<0.05, where a helmet is statistically beneficial. They don't give the rest - which rather suggests that the results were found to be not significant. Yet they don't say that - but they still published that graph. To my mind, this is misreporting. In fact, it's damn close to scientific fraud: by not reporting these numbers they're attempting to make their work look more significant than it actually is.

 

[Wobblers]

Yeah right!

I was hoping that no one had noticed.

 

[benb]

I wouldn't reject this paper for not showing any statistically significant results - I'd reject it for not giving the p-values. Actually, they do give p-values for one injury class - facial injuries, p<0.05, where a helmet is statistically beneficial. They don't give the rest - which rather suggests that the results were found to be not significant. Yet they don't say that - but they still published that graph. To my mind, this is misreporting. In fact, it's damn close to scientific fraud: by not reporting these numbers they're attempting to make their work look more significant than it actually is.

Quite. Withholding of results is a real problem, particularly in pharmaceutical research, but that doesn't mean we should be publishing any paper, no matter how shoddy.

 

[david k]

just because someone who has different view and you don't like doesn't make them preaching about it, it is freedom of speech whether you agree with the subject or not!!

Agreed, but you will find that posting anything but 'anti helmet' is classed as preaching or promoting helmet use, as their only issue (claimed) is compulsion and not helmet use, they have no argument with you unless they can claim you are preaching, hence any helmet wearer/poster will be condemned as a preacher so they can pour scorn on you and continue their circular debates and logic

Edit: some are not as arsey as others but they tend not to post anymore