Milgrom's at it again

[Badly Shaved Monkey]

http://www.liebertonline.com/doi/abs.../acm.2009.0071

When Michael Rawlins, head of NICE, wrote a paper reflecting on the problems of evidence in medicine and, apparently, questioning the role of controlled trials it was inevitable that the woos would try to exploit it.

Our favourite superhero duly obliged by reading it while wearing his special DUllgoggles, or possibly by only reading the press reports.

Now Lionel Milgrom has decided to play the same game. Unfortunately the meat of his article is hidden behind a paywall.

Here is what I posted at sciencebasedmedicine drawn directly from Rawlins' actual paper, none of it can rescue homeopathy from the trashcan, e.g.


I consider historical controlled trials should be accepted
as evidence for effectiveness, provided they meet all of the following conditions:
1 The treatment should have a biologically plausible basis. This is met by all the treatments shown in Tables 4 and 5.
2 There should be no appropriate treatment that could be reasonably used as a control. The term ‘appropriate’ would exclude, for example, the use of bone marrow transplantation as an alternative to enzyme replacement therapy in the treatment of Gaucher’s disease.
3 The condition should have an established and predictable natural history. I prefer this phraseology to ‘poor prognosis’. Conditions such as port wine stains may significantly impair patients’ quality of life without threatening life expectancy.
4 The treatment should not be expected to have adverse effects that would compromise its potential benefits. This has to be a sine qua non.
5 There should be a reasonable expectation that the magnitude of the benefits of the treatment will be large enough to make the interpretation of the benefits unambiguous. A ‘signal-to-noise’ ratio of 10 or more appears to be strongly suggestive of a genuine therapeutic effect.23,25 The magnitude of the ‘signal-to-noise’ ratio representing a ‘dramatic’ (ie 10-fold) response, however, is based on impression and is not (at present) supported by any substantive empirical evidence.”
“Before-and-after designs, in conditions with a fluctuating natural history, are of little value”

[shpalman]

http://www.liebertonline.com/doi/abs.../acm.2008.0200

Quote:

... the idea of mapping sequences of outcomes of experiments to sequences of random variable [sic], which has been sketched in our prior paper, is now presented in more detail. Additionally, the possible responsibility of nonlocal effects similar to those being recognized in quantum mechanics is discussed here in a more comprehensible way.

I really hope this issue will be free to access at some point...

[Badly Shaved Monkey]

Originally Posted by shpalman

... the idea of mapping sequences of outcomes of experiments to sequences of random variable [sic], which has been sketched in our prior paper, is now presented in more detail. Additionally, the possible responsibility of nonlocal effects similar to those being recognized in quantum mechanics is discussed here in a more comprehensible way.

I like the idea that they regard the mapping of their results to random variables is a good thing.

I think I find his ideas perfectly comprehensible. They are the flounderings of someone who can't face the fact that he is wrong.

[shpalman]

Originally Posted by Badly Shaved Monkey

I like the idea that they regard the mapping of their results to random variables is a good thing.

I can't possibly imagine what purpose that would serve. I wonder what paper he's citing.

Meanwhile I notice Milgrom is a coauthor on a proper paper: http://www.rsc.org/publishing/journa...p?doi=b812724g

[Badly Shaved Monkey]

OK, this time the stupid really hurts badly.

Milgrom;

"Weingärtner develops a probabilistic argument based on
the 18th century Swiss mathematician Jacob Bernoulli’s weak
law of large numbers. This allows the problem of random
reproducibility in homeopathy experiments to be tackled, by
mapping series of experimental outcomes onto the convergence
behavior of series of random variables.
Complicated as this might sound, it is rather similar to
working out the odds of heads or tails appearing when one
tosses a coin: Though each coin toss is random, over a large
enough series of tosses, the chance of a head or tail in all
probability converges on a definite number. From this point
of view, it does not matter that reproducibility happens randomly:
All that matters is that the probability of experimental
reproducibility occurring in every homeopathy experiment
in a large enough series of trials will converge on
a definite number."

Yes, Lionel, a series of coin tosses will converge on a definite number for the number of heads or tails, that number is 0.5 the random probability of a single throwing yielding a head or tail.

This does not help homeopathy, it is just another way of stating that since the long-term behaviour of homeopathy trials is to show feck all effect then the effect of homeopathy in one trial is feck all.

[Badly Shaved Monkey]

shpalman, please do the maths again...

http://www.liebertonline.com/doi/abs.../acm.2008.0200

[Badly Shaved Monkey]

Originally Posted by shpalman

I can't possibly imagine what purpose that would serve. I wonder what paper he's citing.

Meanwhile I notice Milgrom is a coauthor on a proper paper: http://www.rsc.org/publishing/journa...p?doi=b812724g

I've been told before that porphyrins were his field before homeopathy set in.

[kellyb]

Who is "O Weingartner" anyway? Homeopathy proponents claim he's a "theoretical physicist" but google scholar only shows him talking about homeopathy, and a couple of references from the 80's on something about "Biophoton Emission" (which sounds like some different woo...anyone know what that is?)

http://scholar.google.com/scholar?nu...:o-weingartner

Since Milgrom likes to tie homeopathy and Biophoton Emission together, I'm pretty sure it's the same guy.

[shpalman]

Originally Posted by shpalman

I can't possibly imagine what purpose that would serve. I wonder what paper he's citing.

Oh, something in German.

[Badly Shaved Monkey]

Originally Posted by shpalman

Oh, something in German.

No, no, it's this one. It's not in German but it is in gibberish and I believe you have some experience of speaking gibberish to loons, so might be able to translate for us.

[Mojo]

Originally Posted by Badly Shaved Monkey

OK, this time the stupid really hurts badly.

Milgrom;

"Weingärtner develops a probabilistic argument based on
the 18th century Swiss mathematician Jacob Bernoulli’s weak
law of large numbers. This allows the problem of random
reproducibility in homeopathy experiments to be tackled, by
mapping series of experimental outcomes onto the convergence
behavior of series of random variables.
Complicated as this might sound, it is rather similar to
working out the odds of heads or tails appearing when one
tosses a coin: Though each coin toss is random, over a large
enough series of tosses, the chance of a head or tail in all
probability converges on a definite number. From this point
of view, it does not matter that reproducibility happens randomly:
All that matters is that the probability of experimental
reproducibility occurring in every homeopathy experiment
in a large enough series of trials will converge on
a definite number."

Yes, Lionel, a series of coin tosses will converge on a definite number for the number of heads or tails, that number is 0.5 the random probability of a single throwing yielding a head or tail.

This does not help homeopathy, it is just another way of stating that since the long-term behaviour of homeopathy trials is to show feck all effect then the effect of homeopathy in one trial is feck all.

Actually, they're getting pretty close. The BHA reviewed the literature up to the end of 2005 a while back, and found that:

Quote:

Our review of 119 randomised peer-reviewed clinical trials (RCTs) of homeopathy to the end of 2005 shows that 49% show positive results for homeopathy. Only 3% were negative. The remaining 48% were inconclusive

Doing the maths, that's 58 positive, 4 negative and 57 "inconclusive".

On their website currently is a FoH document which states:

Quote:

Up to the end of 2008, 136 RCTs had been published: 59 positive; 9 negative; 68 not statistically conclusive.

So in the last three years, that's 17 trials and only one positive result. That's not far off the 5% false positives you would expect.

[Mojo]

BTW, the FoH claims "134 randomised controlled trials (RCTs) have been published to the end of 2007: 59 positive, 8 negative and 67 not statistically conclusive", so 2008 appears to have been rather a quiet year, with a grand total of 2 RCTs published, one negative and one inconclusive.

[Mojo]

Originally Posted by Badly Shaved Monkey

I've been told before that porphyrins were his field before homeopathy set in.

Here's his Persoanl [sic] biography.

[shpalman]

Originally Posted by Badly Shaved Monkey

No, no, it's this one. It's not in German but it is in gibberish and I believe you have some experience of speaking gibberish to loons, so might be able to translate for us.

Originally Posted by Otto Weingärtner

In a previous study [5], we have already tackled the problem of reproducibility for experiments with homeopathic potencies but from a slightly different point of view...

5. Weingärtner O. On the problem of reproducing experiments with homeopathic potencies [in German]. Zeitschr Anomalistik 2007;7:270–285.

[Badly Shaved Monkey]

Originally Posted by shpalman

5. Weingärtner O. On the problem of reproducing experiments with homeopathic potencies [in German]. Zeitschr Anomalistik 2007;7:270–285.

Oh, I see. Wires now uncrossed.

[Soapy Sam]

BSM - I've suggested this before to no response.

Would it be worthwhile for a few interested individuals to pool some funds to subscribe to one copy of certain journals off or online?

Naturally, we should obey all relevant legislative and moral strictures relating to intellectual (and less intellectual) property rights.

I could find £20 a year without breaking the bank.
Anyone else?

[Yuri Nalyssus]

Originally Posted by Soapy Sam

BSM - I've suggested this before to no response.

Would it be worthwhile for a few interested individuals to pool some funds to subscribe to one copy of certain journals off or online?

Naturally, we should obey all relevant legislative and moral strictures relating to intellectual (and less intellectual) property rights.

I could find £20 a year without breaking the bank.
Anyone else?

I think a group doing that would be on very dodgy ground. How could enough of a paper be shared sufficiently to give group members a genuine insight without infringing property rights? Otherwise the group members would just be relying on the actual owner of the subscription/article to interpret for them which wouldn't be very satisfying.

There are a number of free, online, CAM (and other) publications available though, some of which Milgrom has published in - it might be worthwhile compiling a list of them.

Yuri

[shpalman]

I can only think of eCAM: http://ecam.oxfordjournals.org/cgi/c...abstract/4/1/7

[shpalman]

Originally Posted by Badly Shaved Monkey

shpalman, please do the maths again...

http://www.liebertonline.com/doi/abs.../acm.2008.0200

Well I've read it but despite the claims in the introduction that

Quote:

the present paper is concerned with the “presentation of a specific procedure” to obtain reproduction

and

Quote:

the possible responsibility of nonlocal effects similar to those being recognized in quantum mechanics is discussed here in a more comprehensible way

I really don't find anything more insightful than "if you keep doing an experiment the average outcome should converge onto the right answer". Except of course it doesn't seem to be put as clearly as that (until the end of the appendix anyway), because that would be banal. He just makes it look impressive by copying out part of a stats book.

http://en.wikipedia.org/wiki/Law_of_...s#The_weak_law

ETA: and anyway, we shouldn't just be averaging the yes/no outcome of trials of homeopathy to get the probability it works, we should be doing a meta-analysis of the results to take into account how strong the effect was, and how statistically significant it is, in each trial. Weingärtner explains how bigger trials are better than small ones because you get closer to the right answer so he obviously requires that the meta-analysis should be weighted towards trials with more participants. You know, like in Shang et al....

It continues to amaze me how they wrestle with the problem of homeopathy being shown not to do anything.

[shpalman]

... and now I've blogged it: Otto Weingärtner.

[shpalman]

... and now, Cyril W. Smith has decided to share his thoughts on Otto Weingärtner's paper: Can Homeopathy Ameliorate Ongoing Sickness?

(That's the best title he could think of which gives "chaos" as its acronym.)

Quote:

The thesis of this Editorial is that, between the states of health and disease, a patient passes through a state of mathematical chaos (as hidden in the title)... From the clinical and homeopathic point of view, any experiment involving the chaotic domain is nonrepeatable from any given initial condition, even if the underlying theory is deterministic.

No, Cyril, a deterministic system will always evolve the same way from the same initial conditions - the problem with chaotic systems is that slight changes in the initial conditions can end up giving wildly different (as opposed to slightly different) results.

[Mojo]

This suggests that when a homoeopath gives a patient a remedy, they will have no idea what will happen.

[Gord_in_Toronto]

Originally Posted by Mojo

This suggests that when a homoeopath gives a patient a remedy, they will have no idea what will happen.

For various values of Nothing?

[Darat]

Originally Posted by Mojo

This suggests that when a homoeopath gives a patient a remedy, they will have no idea what will happen.

I know - it is quite shocking when they say something that is accurate isn't it?