There's an elegant article at Medscape past times Christopher Labos called "It Ain't Necessarily So: Why Much of the Medical Literature Is Wrong." Key points: Given a statistical association betwixt X together with Y, almost people brand the supposition that X caused Y. However, nosotros tin easily come upwards up alongside five other scenarios to explicate the same situation.
1. Reverse Causality
Given the association betwixt X together with Y, it is genuinely every bit probable that Y caused X as it is that X caused Y.
2. The Play of Chance together with the DICE Miracle
Whenever a written report finds an association betwixt two variables, X together with Y, in that place is ever the possibility that the association was exactly the effect of random chance.
The Frequency of False Positives
It is sometimes humbling together with fairly disquieting to intend that conduct a opportunity tin play such a large purpose inward the results of our analyses.
3. Bias: Coffee, Cellphones, together with Chocolate
Bias occurs when in that place is no existent association betwixt X together with Y, but i is manufactured because of the agency nosotros conducted our study.
4. Confounding
Confounding, dissimilar bias, occurs when in that place genuinely is an association betwixt X together with Y, but the magnitude of that association is influenced past times a 3rd variable.
Real-World Randomization
Confounding tin endure dealt alongside through randomization. When written report subjects are randomly allocated to i grouping or exactly about other purely past times chance, whatever confounders (even unknown confounders) should endure every bit introduce inward both the written report together with command group. However, that assumes that randomization was handled correctly.
5. Exaggerated Risk
Finally, allow us brand the unlikely supposition that nosotros accept a lawsuit where aught went wrong, together with nosotros are gratuitous of all of the problems discussed above. The greatest danger lies inward our misinterpretation of the findings.