Agree with above concerns. Also, there seems to be a greater trend over the years for use of meta-analyses (combining a bunch of studies in one heap in order to increase number of outcomes measured, which should yield more statistically significant results). It's also one way of publishing a paper without the time, expense, or aggravation of setting up or conducting a study of your own. The authors rarely select every study available on the subject, & are often treating very different study methods as equivalent. A meta-analysis will describe its study selection criteria & state potential sources of error up front & in detail. However, a diligent reader who wishes to determine if the results are credible must essentially duplicate most of the authors' data gathering process & actually read the articles they are citing. This can take hours, & is seldom done by people using the information to make decisions. A poorly designed study can live forever in meta-analyses. Will become even more common with AI.
|