Monday, November 11, 2013

The scale of scientific confidence

It is a truism that no scientific result is final.  Scientists are open to their most cherished scientific principles being overthrown.  Nevertheless, any scientist will have a scale of confidence in scientific results.  An interesting question is where on this scale would one place the finding of climate science predicting global warming and climate change compared to say, to some of few results of IQ and genes.



Let me consider for example, the first (and scientifically strongest) result discussed in my post on In My Genes.
Markerrs174575
The researchers found that being breastfed raised a person’s IQ an average of six to seven points, but only among those who had at least one C at rs174575. Among those with a G at both copies of rs174575, breastfeeding appeared to have no effect on intelligence. This result was found in two independent groups, one composed of 1,848 British children and the other of 858 children of European ancestry from New Zealand. The effect associated with this SNP was larger than the one described below for rs1535, which is included for customers who do not have data from rs174575.
The study finds a strong effect, the design of the study seems sound, the samples tested are large, and it is no doubt a fine piece of scientific work.

But!
  1.  As far as I know, no one has repeated the study.  My guess is it will be a while before anyone does - because there is essentially a gold rush out there of new discoveries to be made and no one is going to spend much time on already trodden ground.
  2. The study, good as it is, finds only a correlation, not a causation.  To quote the paper:
The genetic moderation of breastfeeding effects on IQ is unlikely to be directly caused by the analyzed SNP, and the molecular mechanism by which rs174575 may influence cognitive development is not known .
Now, science is all about discovering chains of causes.  Discovering a correlation is only the beginning of science.   Until the causal chain is identified we do not know what knowledge the correlation represents.  As Wiki tries to explain:
For any two correlated events A and B, the following relationships are possible:
  • A causes B;
  • B causes A;
  • A and B are consequences of a common cause, but do not cause each other;
  • There is no connection between A and B; the correlation is coincidental.
Less clear-cut correlations are also possible. For example, causality is not necessarily one-way; in a predator-prey relationship, predator numbers affect prey, but prey numbers, i.e. food supply, also affect predators.
Let us compare this with climate science.   Climate science relies on well-established physical law that describes causes, not just correlations.  A simple back-of-the-envelope calculation (I think CIP has one on his blog somewhere which I can't find, but I did find this pointer to a simple textbook calculation.) shows the main effect of the increasing concentration of carbon dioxide in the atmosphere.  I think even the rotational and vibrational levels of the carbon dioxide molecule that make it a "greenhouse gas" are accessible to computation from first principles in quantum mechanics.

The great difficulty in climate science is that the actual land-atmosphere-ocean-ice is a complicated system and applying physical laws without using approximations is computationally intractable.  The finer details of climate change may always be beyond our grasp.   But there are numerous, independent  peer-reviewed climate models that establish the main results of climate science.   Both the computations and the measurements of the Earth have greatly improved over the past couple of decades. 

To put it another way, to overthrow the results of climate science will take an enormous reworking of our understanding of how to apply physical laws to the Earth.  It would mean that a large number of people have made a large number of mistakes.  In contrast,  it will take just one contradictory study to throw doubt on the genetic study that I mentioned.

Clearly the confidence we ought to have in the results of climate science is much, much higher than we should have in things the results of the genetic study I cited.   Indeed, I would call into question the scientific judgment of anyone who put both results at the same level of confidence.