LikeLike

]]>LikeLike

]]>LikeLike

]]>LikeLike

]]>LikeLike

]]>Memories are short in alalalarmist land.

LikeLike

]]>LikeLike

]]>Like you I’ve always had reactions against statistics so spent much of my 2nd year at University (Maths major) plaguing my stats tutor with ‘proofs’ it wasn’t ‘proper maths’. 47 years later, I can still remember his exasperated sighs of “but it works…” and glum face.

Statistics starts with using probabilities for independent events acting on independent variables at indetermined times e.g. tossing a perfectly balanced coin with identical tosses over and over again. You can probably see the most obvious problem for surface temperatures – they’re time-dependent. There are ways around it, and theories about the best approaches to use. But I shake my head in disbelief when I responses from climate scientists along the lines of “We’ve used it before, and it turned out all right”. Does this mean they predicted something that happened? Or that predicted what they expected to find? Or that everyone else thinks it’s OK? And so on.

So, the 1st 2 things stats users need to do is (i) ensure variables are independent and (ii) arrange to take samples to avoid the events and variables having any commonality.

And secondly, always remember correlation says absolutely zilch about causation.

I like the example of carbon-dioxide. We know that a warming sea will transpire carbon-dioxide to the atmosphere, therefore we expect to find a correlation between increases of temperature and carbon-dioxide – with a time-lag (CO2 later than Temp) of indeterminate size. We do, over the long-term.

LikeLike

]]>Thanks!

Richard

LikeLike

]]>