Friday, November 21, 2008

RNLI - Saving lives at sea, funding quackery on land

The Royal National Lifeboats Institute (RNLI) is the charity that saves lives at sea.

As required by the Charity Commission, it defines its objects as:
FIRSTLY ... TO SAVE LIVES, PROMOTE SAFETY AND PROVIDE RELIEF FROM DISASTER AT SEA AND, SECONDLY, TO SAVE LIVES, PROMOTE SAFETY AND PROVIDE RELIEF FROM DISASTER ON INLAND WATERS


When I'm not laid low I like to go scuba diving and instructing around the UK coast and have one occasion been involved in an RNLI operation. They are an amazing service and save myriad lives per year through sheer bravery by volunteers. The Charity Commission webpage on the RNLI shows them to have 42400 volunteers compared with 1332 staff. I feel it is my duty to donate to them on a regular basis, given that I may need their assistance one time in the future.

What I don't want them to do is to spend any money funding quackery - it turns out that they appear to have been funding a 'trial' into the homeopathic treatment of weaverfish stings.

Weaverfish are a small spiny-backed fish that annoyingly like to hang around just under the sand in shallow water with their fins sticking up. Weaverfish snacks and small children end up getting stung which can bring up to 30 mins of pain, which is remedied against by placing the sting site in hot water.

Now, the Cornwall College Camborne are investigating as to whether a 200c preparation (remember that's diluted 1:1,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,
000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,
000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,
000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,
000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,
000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,
000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,
000 or 1:10^400) of bee sting given in tablet form would have any effect. The homeopathic treatments were provided by Helios Homeopathy (Avid readers will no doubt remember that Helios were quite happy to sell homeopathic remedies and prophylactics for Malaria - potentially fatal, of course - without a shred of evidence to show efficacy, because none exists. The MHRA have since forced them to remove the offending products from their shelves).

The website links to any results for 2006 and 2007 give you a 'bad link' result, so I tried phoning a few people. Firstly, RNLI (BeachGuards) PR HQ - they had never heard of the study, secondly the chap who was running the 'trial', David Retford, was not for answering his phone. His webpage in the College Cornwall website does have some results though, from 2007.

We find out that n=24 for 2007 with 59 in 2006, so at any rate the result is not worth a whole lot. His results state:
the time to it took for participants to leave was less in group A: an average of 18.19 minutes (homeopathic group), as opposed to 19.71 minutes in group B (placebo group).


He claims that the 'rate of pain relief' is significant at p=0.051. I claim that with n=24 the whole thing is pointless and one and a half minutes is neither here nor there.

As with the Natural History Museum getting involved with Homeopathic databases, the RNLI is lending its very good name to be associated with quackery who will no doubt wear the badge with pride. The RNLI is there to save lives (their own mission statement) not to fund quackery.

David Retford also requests:
We need further funding from both the College and the RNLI to continue.

I will be writing to the RNLI and to try and make sure this doesn't happen.

Thanks to David Barratt and other Bad Sciencers for the heads up on this.

BPSDB

10 comments:

  1. Bill S. Preston Esq.November 21, 2008 at 2:04 PM

    *cough cough ahem harrumph*

    1:1^400 is actually a pretty strong concentration, homeopathy-wise. Feel free to delete this comment and preserve your modesty...

    ReplyDelete
  2. No idea what you're talking about ;)

    (fixed - and ta!)

    ReplyDelete
  3. Since when was p=0.051 significant?

    ReplyDelete
  4. Exasperating stuff. A single penny that is not spent on maintaining and promoting the service they provide is a penny embezzled in my view.

    ReplyDelete
  5. Anonymous - indeed, my knowledge of them thar stats is not brilliant, but I would have thought you put the p value at which you would the reject the null hypothesis (i.e that homeopathic bee sting is no different from placebo) and then do the calculation to find out what the result was. p=0.051 means (if you'd chosen 5% level) that you would accept the null hypothesis. I.e Homeopathy no different from placebo.

    You know what, I could have told them that before they started, without wasting money on quackery which was previously destined for lifeboats.

    ReplyDelete
  6. My stats are also a bit flaky, I’m in a hurry and doing this on the fly so take the following with a pinch of salt but in addition to the one-tail two-tail thing:

    1. The z test assumes a normally distributed data. If you take just a few measurements from a normal distribution you are more likely to get measurements near the mean than far away from the mean. Thus if you do a z test with low replicate numbers you will under estimate the true variance of the data. This will in turn lead to an under estimate of the significance value. This is why you use a t-test for low replicate numbers as it takes this effect into account via the degrees of freedom. Low replicate numbers means less than 30 per group.

    2. He is measuring a “time to event” variable. Such data do not conform to normal distributions and require completely different handling that is beyond my knowledge.


    So he was wrongly applied a statistical test that was never appropriate to his data in the first place. The results are just totally meaningless in that we don't know if the data support regection of the null hypothesis or not.

    Epic Fail.

    ReplyDelete
  7. Yes, the stats sound a bit bodged...

    Not to mention the enormous "prior probability" that homeopathy does seven shades of f*!k all, of course...

    ..so marginal statistical anything ain't gonna cut it. Remember James Randi and the famous horse-vs-unicorn discussion (see the last couple of paras here).

    Like you, I am sorry to hear of the RNLI subsidising this nonsense. When I was a nipper I had a great regard for the RNLI and used to expend considerable effort as a primary school kid trying to get my class to make them our annual class charity. Sadly I tended to lose out to the RSPCA or PDSA, which left me with a lifetime distrust of fluffy ickle kitty-cats and their human friends.

    ReplyDelete
  8. Hmmm, really they should have done a randomisation test, which is both non-parametric and uniformly most powerful. Has anyone got the original numbers? If so, I'll do it for them. Not, of course, that it's likely to be anything but utterly indecisive (as always)

    Duchy College, does suggest an influence of the Quacktitioner Royal. Does anyone know if that is the case?

    Having had a boat foe much of my life, I've always been strongly for the RNLI, so this is very disappointing. A stiff letter has gone to them.

    At least as interesting, though, is the antics of ITEC and OfQual,who 'validate' this sort of stuff. They merit more investigation. The FoIs have gone off.

    ReplyDelete
  9. I think you should use 2 tail instead of one tail though I did not properly understand what the idea it wanted to convey but thanks for sharing it with us.

    ReplyDelete
  10. DC - thanks for comment and for the link on your blog.

    The off-the-wall OfQual and Edexcel-OUT story is absolutely mind-boggling. I've been through the story and can't really decipher it - how can they operate in such a muddle?

    All the best with the FoIs - this story merits the full force of DC & Associates!

    Moshe - I think you are right, but my statistics are rusty, possibly beyond repair!

    ReplyDelete

Share it