3 ms·
It is very easy to get bad results in medical studies. Placebo effects are surprisingly powerful, and you need double-blinding to get rid of them. Beyond that,
by jcranmer 8y ago
It is very easy to get bad results in medical studies. Placebo effects are surprisingly powerful, and you need double-blinding to get rid of them. Beyond that, you have the typical slew of statistical fraud that can be committed: ignoring bad results (meaning that published results are more likely to be statistically flukes themselves), p-hacking, insufficient sample size, etc.
A proper drug approval trial process is one of the best ways to get statistically valid results. It is also horrifically expensive--you're looking at >$100 million for a phase III trial, whether or not you succeed (this is the main reason drugs are so expensive: you're paying for phase III failures). The studies you see are generally of much poorer quality, and most people who are interested in the subject tend to have a result they are aiming for, which increases the risks of statistical fraud greatly.