4 ms·
If you don't understand that correlation doesn't imply causation, you should be sent back to your first year of high school and forced to try it all again.
by M_Grey 10y ago
If you don't understand that correlation doesn't imply causation, you should be sent back to your first year of high school and forced to try it all again.
- pmoriarty 10y agoWhat, if not correlation, does imply causation?
- M_Grey 10y agoI'm sorry, is that an incredibly lazy way of arguing that correlation does imply causation, but you feel like reversing it and dumping the burden of proof onto me? No.
- pcrh 10y agoCorrelation suggests causation, but does not prove it.
- M_Grey 10y agoIt suggests correlation; get a big enough dataset and you'll find correlations all over the place. They don't imply a causal relationship, at all. http://imgs.xkcd.com/comics/heatmap.png http://imgs.xkcd.com/comics/heatmap.png
- pcrh 10y agoIndeed, correlations are not sufficient to establish cause. However... things that do cause a specific effect also tend to correlate. So correlations are a basis for hypotheses that can be experimentally tested. To wit, the hypothesis presented in the post linked to by OP can be tested by treating people with Alzheimer's disease with antimicrobial agents. Unfortunately, in what concern people (disease and dying) experiments on humans (aka clinical trials) are expensive, with uncertain returns. So the experiments that could test hypotheses concerning Alzheimer's disease in real actual people are rarely performed to the level statistically needed. Instead we get a plethora of studies in mice with "suggestive"implications for what might occur in humans. Along with random studies that are not worth more than anecdotes (see those I commented on above).