## Wednesday, April 18, 2007 ... /////

### Richard Lindzen: Iris effect remains alive and well

Some readers may want to look at

about the Iris effect, originally published in 2001. The claimed agreement of existing climate models with reality may be described as curve fitting. There are good reasons to think that there is a couple of important negative feedbacks related to water vapor and clouds that make the CO2 climate sensitivity small.
ABC of the iris effect: when the Earth is warming, the rain at places where air flows in the direction up becomes more intense. This reduces the amount of water droplets available for high-altitude cirrus clouds. Warming thus reduces the number of cirrus clouds and because these clouds have a warming effect, the overall impact of this mechanism is a slight cooling which means that the ultimate warming will be smaller than if the iris effect wouldn't exist.
A 2007 paper by Roy Spencer et al. on cirrus clouds brought some new evidence for the iris effect. See also
Roy Spencer: Global warming and Nature's thermostat: precipitation systems
Incidentally, NOAA climate models have just obtained the same result about the future hurricane rate
as people who have read at least initial chapters of meteorology textbooks know - although the microscopic details of the explanations are different. (Thanks to Alexander Ač, no kidding.)

You may also read a semi-popular text of a lecture by Prof Lindzen called
about the ways how bias is being introduced to climate science. When Lindzen talks about a "certain kind of errors" that are selectively "corrected", we may think of many examples. Unfortunately, a recent correction of
may be another example. The file above contains the correction (first 5 pages) as well as the original paper (last 15 pages). The authors originally reported that the ocean lost 32+-11 zettajoules of heat (zetta is 10^{21}, after tera, peta, exa) between 2003 and 2005 (it cooled down). In the correction, they don't say what the new numbers and error margins are but they offer a lot of "biases" whose "correction" will probably change their cooling into warming, or at least reduce the cooling.

Given the fact that zetta- doesn't appear in the original paper and the correction systematically misspells zetta- as zeta-, one may conjecture that the correction to the paper was written by a less intelligent but more concerned writer than the original paper.

It's very plausible that there can exist errors and biases but if one looks for them with a particular "big idea" in mind, it is likely that he will get skewed results. And it is hard to avoid the feeling that skewed results are actually the real goal in many cases.

And that's the memo.

#### snail feedback (4) :

not really relevant but still quite funny: Richard Lindzen's views on the health risks of smoking (in wikipedia.org).

regards,

j

I don't know why you find it "funny". I think that he is very right. Concerning second-hand smoking, which is really the thing that he has spoken about publicly, it is a very vaguely documented correlation.

Take this: spouse smoking exposure increases cancer risk by 20%. This is supposed to be a great evidence for the link. The numbers are statistically insignificant. The estimated ratio of cancer rates for spouses of smokers vs. complete nonsmokers is between 1.01 and 1.37 at 95% confidence level.

There are other numbers of this type and it just happens that the lower rate is always 1.01. Surely a coincidence, right? At high confidence level, they always want to stay above one.

With huge errors, it is simply completely plausible to think that second-hand smoking doesn't influence the lung cancer rate at all. What they have is a 2-sigma effect, and one may think that even this effect was caused by a bias and predetermined goals of the research. I am almost completely unconvinced by the actual numbers.

Unlike Richard Lindzen, I hate cigarettes. But the numbers make it pretty clear that he has a very strong point.

Personally, I think that there is evidence that first-hand smoking significantly increases lung cancer rates - 80% of lung cancers can be correlated with smoking although only 50% of people smoke - but I don't think that the people who think otherwise are completely crazy.

dear lubos,

i found the quote "He'll even expound on how weakly lung cancer is linked to cigarette smoking. He speaks in full, impeccably logical paragraphs, and he punctuates his measured cadences with thoughtful drags on a cigarette." rather amusing...

but i definitely don't want to open this can of worms and argue about these things as well (although i should be honored that you only seem to ignore my comments pertaining to climate related issues;-), and hence will just comment in general.

medical studies often contradict each other. results claiming to have "proven" some causal connection are confronted with results claiming to have "disproven" the link, or vice versa. this dilemma affects even reputable scientists publishing leading medical journals. the topics are divers:
- high-voltage power supply lines and leukemia [1],
- salt and high blood pressure [1],
- heart diseases and sport [1],
- stress and breast cancer [1],
- smoking and breast cancer [1],
- praying and higher chances of healing illnesses [1],
- the effectiveness of homeopathic remedies and natural medicine,
- low frequency electromagnetic fields and electromagnetic hypersensitivity [2],
...

basically, this is understood to happen for three reasons:
i.) the bias towards publishing positive results,
ii.) incompetence in applying statistics,
iii.) simple fraud.

i.)
---
publish or perish. in order the guarantee funding and secure the academic status quo, results are selected by their chance of being published.

an independent analysis of the original data used in 100 published studies exposed that roughly half of them showed large discrepancies in the original aims stated by the researchers and the reported findings, implying that the researchers simply skimmed the data for publishable material [3].

this proves fatal in combination with ii.) as every statistically significant result can occur (per definition) by chance in an arbitrary distribution of measured data. so if you only look long enough for arbitrary results in your data, you are bound to come up with something [1].

ii.)
-----

often, due to budget reasons, the number of test persons for clinical trials are simply too small to allow for statistical relevance. [4] showed next to other things, that the smaller the studies conducted in a scientific field, the less likely the research findings are to be true.

statistical significance - often evaluated by some statistics software package - is taken as proof without considering the plausibility of the result. many statistically significant results turn out to be meaningless coincidences after accounting for the plausibility of the finding [1].

one study showed that one third of frequently cited results fail a later verification [1].

another study documented that roughly 20% of the authors publishing in the magazine "nature" didn't understand the statistical method they were employing [5].

iii.) a.)
------------

two thirds of of the clinical biomedical research in the usa is supported by the industry - double as much as in 1980 [1].

it was shown that in 1000 studies done in 2003, the nature of the funding correlated with the results: 80% of industry financed studies had positive results, whereas only 50% of the independent research reported positive findings.

it could be argued that the industry has a natural propensity to identify effective and lucrative therapies. however, the authors show that many impressive results were only obtained because they were compared with weak alternative drugs or placebos. [6]

iii.) b.)
------------
quoted from wikipedia.org:

"Andrew Wakefield (born 1956 in the United Kingdom) is a Canadian trained surgeon, best known as the lead author of a controversial 1998 research study, published in the Lancet, which reported bowel symptoms in a selected sample of twelve children with autistic spectrum disorders and other disabilities, and alleged a possible connection with MMR vaccination. Citing safety concerns, in a press conference held in conjunction with the release of the report Dr. Wakefield recommended separating the components of the injections by at least a year. The recommendation, along with widespread media coverage of Wakefield's claims was responsible for a decrease in immunisation rates in the UK. The section of the paper setting out its conclusions, known in the Lancet as the "interpretation" (see the text below), was subsequently retracted by ten of the paper's thirteen authors.

[...]

In February of 2004, controversy resurfaced when Wakefield was accused of a conflict of interest. The London Sunday Times reported that some of the parents of the 12 children in the Lancet study were recruited via a UK attorney preparing a lawsuit against MMR manufacturers, and that the Royal Free Hospital had received £55,000 from the UK's Legal Aid Board (now the Legal Services Commission) to pay for the research. Previously, in October 2003, the board had cut off public funding for the litigation against MMR manufacturers. Following an investigation of The Sunday Times allegations by the UK General Medical Council, Wakefield was charged with serious professional misconduct, including dishonesty, due to be heard by a disciplinary board in 2007.

In December of 2006, the Sunday Times further reported that in addition to the money given to the Royal Free Hospital, Wakefield had also been personally paid £400,000 which had not been previously disclosed by the attorneys responsible for the MMR lawsuit."

wakefield had always only expressed his criticism of the combined triple vaccination, supporting single vaccinations spaced in time. the british tv station channel 4 exposed in 2004 that he had applied for patents for the single vaccines. wakefield dropped his subsequent slander action against the media company only in the beginning of 2007. as mentioned, he now awaits charges for professional misconduct. however, he has left britain and now works for a company in austin texas. it has been uncovered that other employees of this us company had received payments from the same attorney preparing the original law suit. [7]

epilog
-----------

should we be surprised by all of this? next to the innate tendency of human beings to be incompetent and unscrupulous, there is perhaps another level, that makes this whole endeavor special.

the inability of scientist to conclusively and reproducibly uncover findings concerning human beings is maybe better appreciated, if one considers the nature of the subject under study. life, after all, is an enigma and the connection linking the mind to matter is elusive at best (i.e., the physical basis of consciousness).

the bodies capability to heal itself, i.e., the placebo effect and the need for double-blind studies is indeed very bizarre. however, there are studies questioning, if the effect exists at all;-)

best regards,

james

---

[1] this article in the magazine issued by the Neue Zürcher Zeitung by Robert Matthews, who gets the brunt of your rant in this post

[2] C. Schierz; Projekt NEMESIS; ETH Zürich; 2000

[3] A. Chan (Center of Statistics in Medicine, Oxford) et. al.; Journal of the American Medical Association; 2004

[4] J. Ioannidis; "Why Most Published Research Findings Are False" ; University of Ioannina; 2005

[5] R. Matthews, E. García-Berthou and C. Alcaraz as reported in this "Nature" article; 2005

[6] C. Gross (Yale University School of Medicine) et. al.; "Sope and Impact of Financial Conflicts of Interest in Biomedical Research "; Journal of the American Medical Association; 2003

[7] H. Kaulen; "Wie ein Impfstoff zu Unrecht in Misskredit gebracht wurde"; Deutsches Ärzteblatt; Jg. 104; Heft 4; 26. Januar 2007.