2 ms·
One reason its such a problem is that many studies are observational, rather than experimental. There's a lot more room for ex post hypothesizing (first look at
by netcan 7y ago
One reason its such a problem is that many studies are observational, rather than experimental. There's a lot more room for ex post hypothesizing (first look at the data, then pose a hypothesis), which invalidates statistical confidence.
- inlined 7y agoThis reminds me of a story that I unfortunately forget the citation of: The military wanted to conduct an experiment to see if urban or rural recruits would have an easier time matriculating. They felt the hypothesis was obvious: rural recruits would be more “rugged” and handle extreme environments better. The experiment showed the opposite—urban recruits were matriculating better. Suddenly the conclusion was “obvious” as well: urban recruits were more accustomed to rigid lifestyles and could handle the orders barked at them. The lesson is that if two opposite results could be “obvious” then neither are. Observational studies can very easily fall into this trap.
- netcan 7y agoThat story is very close to something Karl Popper could have said. He would have called it pseudo-science.