One of the Fallacious Arguments in which it is asserted because a correlation is demonstrated between A and B, A must therefore cause B. Related to Post Hoc Ergo Propter Hoc (which seems to imply a time sequence between A and B).
Fallacious because:
B could cause A
There might be a C which is a cause of both A and B
A and B could be coincidental; though with a statistically valid set of measurements this is rare.
In statistics, it is known that correlation does not imply dependence in the general case (mathematicians speak of events or processes as being dependent if there is some causal relationship between them, or among them and some other event(s) or process(es)). However, there are many cases (such as two random processes which both have Gaussian distributions) where correlation between the two processes does imply dependence.
In most rhetorical arguments, this statistical observation does not apply--either the things being discussed cannot be measured adequately, or there isn't a statistically valid sample.
This isn't entirely fallacious when utilized with abduction. All determination of causation of real-world events is through empirical evidence. However, it certainly isn't deductive.
An old joke concerns a scientist from the University of Elbonia (usually, the joke is phrased as an ethnic joke with some real nationality the butt of the joke--but I'll borrow the mythical country of Elbonia from Dilbert to keep this Politically Correct) who performs research on flies.
He anesthetizes one fly, puts it on his laboratory table. When the fly comes to, he tells it to fly and pounds on the table. The fly flies away.
He anesthetizes a second fly, removes one wing, and puts it on his table. When the fly comes to, he tells it to fly, and pounds on the table. The fly jumps around a bit, clearly unable to fly with only one wing.
He repeats the experiment with a third fly, removing both wings. The fly, when told to fly, just sits there.
Scientist writes the following conclusion in his journal:
Removing both wings from a fly makes it deaf.
A and B could be coincidental; though with a statistically valid set of measurements this is rare
Actually, this is not that rare, which is why this is potentially an extremely dangerous practice. With a large set of data, it is quite usually quite easy to find statistically valid yet meaningless correlations. This is why one must use one set of data to generate a hypothesis, but a completely independent set of data to validate hypothesis.
Example:
Ice-creams cause murder. When ice-cream sales go up, the murder rate goes up, ignoring the fact that they could both be linked to the temperature... -lrandall
See original on c2.com 