LV= cycle accidents increase

[Nortones2]

There was a puff by LV that implied cycling accidents had rocketed. Ben Goldacre of Guardian "Bad Science" column has looked at the story http://tinyurl.com/davnbs


Fascinating bit to me is that the "research" is described by LV as "PR reviewed scientific evidence" A roundabout way of saying its just crap...

 

[Lurker]

".... So it seems accidents have gradually gone down by a third over 10 years, but LV insurance and YouGov - using something my colleague Charlton Brooker has helpfully termed "PR-reviewed scientific evidence" - say accidents have gone up by a third in just six months, using data that forgets to account for the usual seasonal variation, and seems to get the prevalence of both cycling and accidents wrong, by an order of magnitude...."

No, it's not LV that have described their data in this way, but Charlie Brooker, the Guardian columnist. Here's his profile: www.guardian.co.uk/profile/charliebrooker

Agree with your conclusion, though.... ;o)

 

[downfader]

Both Charlie Brooker and Goldacre are very good columnists. Both seem to use quite good logic and reasoning and have challenged many myths rather well imo. Very worth of back reading imvho

 

[HJ]

I wonder if they could if they could take a look at the "evidence" for the efficacy of cycle helmets, more "PR-reviewed scientific evidence" I suspect, there is a lot of money at stake after all.

 

[byegad]

As a Statistician in a previous life, I have a healthy scepticism of expert's statistical 'proof'.
With careful choice of Data it is possible to 'prove' almost anything!

 

[beanzontoast]

As a Statistician in a previous life, I have a healthy scepticism of expert's statistical 'proof'.
With careful choice of Data it is possible to 'prove' almost anything!

Can you prove that?

 

[Lazy-Commuter]

I've been a long time reader of Ben Goldacre's blog .. he's always informative.

I do the like the bit in the original press release where it says "1 in 20 (6%) admitted to cycling .. " - erm 1 in 20 = 6% !!??

Think that pretty much sums it up really ...

 

[Tynan]

rounding

 

[Lazy-Commuter]

Still would have been better to say 1 in 16 or 1 in 17. There was plenty wrong with the study, but that was just a silly one that made me chuckle.

 

[Arch]

rounding

What, rounding from 5 up to 6? 1 in 20 is hardly a frational percentage...

 

[Lazy-Commuter]

I suspect they ended up with 5-point-something percent and just called it 1 in 20 due to the 5% bit, but then rounded it to 6% in the brackets. I wondered if they'd said "more than 1 in 20" in the text and I missed it, but the copy of the article on Goldacre's blog says "1 in 20 (six per cent)".

I'm sure Tynan's right in that it's just them rounding, albeit in a crazy way, but for me that's kind of the point: they're making such a big thing out of these (dubious) stats and they've got an obvious silly like that in there.