3 ms·
Much of bad “science” can be traced to innumerate psychologists typing their data into an R program, getting a p-value out, and then wildly misinterpreting that
by leephillips 6y ago
Much of bad “science” can be traced to innumerate psychologists typing their data into an R program, getting a p-value out, and then wildly misinterpreting that. They don’t know what a p-value is or what it means (not much, usually). What is a citizen, interested in keeping up with research results, to do? Remember what science is. Science is not searching for correlations in data. If you see a headline about a result that seems important, you can, you must, ignore it until you look at the research paper. It doesn’t matter if it’s not your specialty. Look for hypothesized mechanism and for successful prediction. If it’s not there, none of the reported associations or “statistical significance” mean anything.
- aeternum 6y agoStats and correlations can still be useful. Quantum mechanics is a great example: we can use it to make the most accurate predictions in the history of the world yet most physicists would agree we have no clue as to the mechanism. As I see it, the ecosystem around research journals is mostly to blame. Throughout history scientists recorded both surprising and unsurprising experimental results. Both were treated as important. Now we basically discard and forget experiments / work that is not 'novel'. The bias introduced by doing this is what makes those p-values questionable. Furthermore, for some it seems that the goal has become publication rather than the search for truth.