News for hackers. Stuff that matters.
Re:Massive 15-year Study finds no link between cell phones, ... by mikeash on Friday October 21, 2011 @08:12PM (#3142311)
Correlation does not imply causation, but lack of correlation certainly does imply lack of causation.
Re:Massive 15-year Study finds no link between cell phones, ... by hugh3 on Friday October 21, 2011 @08:22PM (#3142335)
Not always. For instance, the temperature in my fridge is a pleasant four degrees celsius constantly, but the compressor keeps on switching on and off.

In the absence of a feedback loop, though... yes.

Re:Massive 15-year Study finds no link between cell phones, ... by mikeash on Saturday October 22, 2011 @07:36AM (#3143355)
Eh, that's just a lack of detail. With a sufficiently precise thermometer, you'd see the temperature rising when the compressor is off, and falling when it's on.
Re:Massive 15-year Study finds no link between cell phones, ... by TeMPOraL on Saturday October 22, 2011 @09:36AM (#3143483)
Yes, but in real life our data usually isn't infinitely accurate.
Re:Massive 15-year Study finds no link between cell phones, ... by mikeash on Saturday October 22, 2011 @08:27PM (#3144712)
Right, but that just means that you're only able to disprove correlation to within a certain accuracy. Correlation below the detection threshold could still exist, and could still be due to causation. Correlation above the detection threshold does not exist, and rules out causation above the detection threshold.
Re:Massive 15-year Study finds no link between cell phones, ... by gregwebs on Monday October 24, 2011 @04:57PM (#3151545)
This link helps explain why lack of correlation does not imply lack of causation: http://blog.cholesterol-and-health.com/2011/01/lack-of-corre...
Re:Massive 15-year Study finds no link between cell phones, ... by mikeash on Tuesday October 25, 2011 @05:02PM (#3155866)
That explains why lack of detection of correlation does not imply lack of causation. But all of the possible failures that it outlines are failures in detection, not an actual lack of correlation. It's a subtle distinction, but I think it's an important one.
Re:Massive 15-year Study finds no link between cell phones, ... by gregwebs on Tuesday October 25, 2011 @09:43PM (#3156750)
That is a good point - the most accurate statement would be: lack of correlation cannot imply there is no relation.

I think we risk oversimplification by referring to it solely as a failure of detection. The first 3 points of the article mean we could probably adjust the experiment right now (or should have before the experiment began). The 4th point, while technically a failure of detection, could be due to unknown confounders (or as Rumsfeld would say - unknown unknowns). So it could be a failure in detection or a failure in knowledge.

Re:Massive 15-year Study finds no link between cell phones, ... by mikeash on Tuesday October 25, 2011 @10:34PM (#3156880)
My terminology may not be the best. But to put it simply, if you do an experiment and that experiment says that there is no correlation, then to within the limits of your knowledge that there is no correlation, you can also say that there is no causation. Whether those limits are based on your experimental design, external factors, or what have you, is a separate (but completely interesting) question.
Re:Massive 15-year Study finds no link between cell phones, ... by gregwebs on Wednesday October 26, 2011 @10:54AM (#3158824)
I think the key to terminology here is to make the statement in a negative fashion: knowledge is always limited so we can never be certain there is no relation.

It is interesting in this case to try to make the claim that because there is no correlation of cancer to proximity of cell phone use that cell phones probably don't cause cancer in nearby locations in the body. I think we have enough knowledge to make that case, but I might just be arrogant.