Noise and Stuff

Separating signal from noise is a vexed question in science. In the case of man-made global warming climate disruption climate change, this is especially difficult due to there being so many confounding factors: insolation levels, long and short term solar cycles, sun-spot activity, gamma rays, long and short-term orbital variations, clouds, diurnal and annual temperature cycles, volcanic activity, Pacific Decadal Oscillation, Atlantic Multidecadal Oscillation etc. Not to mention: data fudging, data smoothing, woefully inadequate models, adjustments, confirmation bias, publication bias and outright fraud. All this stuff; ignored by Stuff.?Quote.

Stuff accepts the overwhelming scientific consensus that climate change is real and caused by human activity. We welcome robust debate about the appropriate response to climate change, but do not intend to provide a venue for denialism or hoax advocacy.End quote.

How is real science done?

In any experiment comparing some phenomenon A to another phenomenon B, researchers construct two hypotheses: that “A and B are not correlated,” which is known as the null hypothesis, and that ?A and B are correlated,? which is known as the research hypothesis. In science, observation trumps theory (except in climate science). But observation of small signals buried in noise and determining their significance, presents difficulties. Computationally, a significant signal depends on meeting an exceptional combination of processing criteria.

Back to our old friend, the bell-shaped curve.

So, if we examine a noisy set of signals in terms of deviation from a nominal zero or null for the correlation criteria, 67% of observations will fall within 1 ? (sigma) and 95% within 2 ? (sigma is the standard deviation). These are confidence levels. Well, actually, no. 95% confidence means that there is a 1 in 20 chance that the observed correlation is a random occurrence with no attributable cause. Or that it is attributable to some other factor other than the one postulated and being tested for. Now that?s good enough for social (and climate) scientists to claim ?the science is settled? or ?a scientific consensus exists?.

Real scientists apply a much stricter test ? 5 ? (five sigma). That?s not one in 20, it?s one in 3? million. Wow! Does that conclusively prove the hypothesis? No, it merely indicates that, given the hypothesis, there is a one in 3? million chance that the observation was without cause ? a fluke. But there may well be other hypotheses which will give a similar level of confidence with the same exceptional dataset.

We all know that if you mix red and yellow paint you get orange. Let?s have a ?thought experiment? about a universe where science has discovered yellow particles but not red particles.

A deep-thinking scientist theorised (his hypothesis) that red particles might exist. He hypothesised that bombarding stuff with yellow particles might produce an orange particle and thus the presence of an orange particle would prove the existence of the elusive red particle, even if the red particle could not be seen or measured on its own.

After bombarding many materials with streams of high energy yellow particles the sensitive instrumentation finally detected an orange particle. The scientist?s theory was proved correct. Maybe. In our example, ?the presence of orange has no relation to the existence of red? is the null hypothesis.

But how would the scientists know that this was not a random fluke or an aberration in the instrumentation?

After repeating the experiments, the data were analysed and this established to a five-sigma level that the orange particle outcome was indeed due to the presence of a red particle. In other words, you would need repeat the experiments 3? million times before you might find a similar result if the universe did not contain the red particle. Your intrepid researcher may then devise a similar experiment with blue particles in the hope of observing a magenta one.

Incidentally, there are everyday occurrences of odds much greater than one in 3? million: winning the lottery, being killed in a commercial air crash and Stuff doing real journalism.

Thanks to ToBoldLeeGoh for his contribution to this post.