Michael Cox wrote:
I'm not arguing against the pattern just pointing out the need to be careful equating correlation with causation.
People more likely to die in hospital than anywhere else.
Banning hospitals will reduce the death rate.
Michael Cox wrote:There is a fallacy in statistics to do with measuring too many variables then pointing to one anomaly. It is equivalent to getting a machine gun, firing it at a wall, examining your bullet spread to find three holes close together and then drawing a circle around them and declaring "bulls-eye". The pattern fits but doesn't have meaning. Likewise there are major issues associating correlation with causation - correlations don't necessarily have any meaning.
We can point to ANY pair of variables that have both increased over the past 20 years and say look, here is a correlation. Soy consumption has increased. Mobile phone use has increased. Conversely you could point to decreasing cigarette consumption and increasing ceoliac disease and claim cigarettes are preventing ceoliacs.
Nick Kitchener wrote:Agreed. The correlation between swales and ground water recharge doesn't necessarily mean causation... ;-P
Michael Cox wrote: these figures are beyond my personal resources to find but determined/funded researchers should be able to gather it.
Michael Cox wrote:Ah, but there is a big difference...
Observation of correlation -> Hypothesis -> experiment -> conclusion
With the initial proposal in this thread we went
Observation of correlation -> causation