3 ms·
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 refer
by gregwebs 15y ago
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.
- mikeash 15y agoMy 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.
- gregwebs 15y agoI 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.