Hacker News new | past | comments | ask | show | jobs | submit login

Much of the solution to this is straight forward. Methodologies should be clearly expressed in detail at the start of the paper; all studies should be publicly logged before they start (so it's not possible to hide studies which say the "wrong" thing); resources should be made available so factors such as randomisation can be properly understood and executed with little or no additional effort. The more information is available the more likely

I'd go on but Ben Goldacre has said it all far better than I could, and in far greater detail in his book Bad Science which I think has just been released in the US (http://www.amazon.com/Bad-Science-Quacks-Pharma-Flacks/dp/08...).

What is important though is not to assume that because the evidence for conventional medicine is sometimes weak, that that in some way makes the case for less mainstream alternatives.

Flaws in one do not in any way strengthen the other, and for alternative medicine the evidence and studies are almost always either even weaker or non-existent.




...all studies should be publicly logged before they start (so it's not possible to hide studies which say the "wrong" thing);...

Not just logged, but accepted by journals. I.e., you submit a paper, explain the experiment design, and put in dummy tables. The conclusion section is unwritten.

The journal accepts or rejects based solely on methodology - they can't reject you after the fact for going against conventional wisdom or getting a null result.


Pre-publication registration of trials has happened. The major journals agreed they would only publish trials that were registered prior to publication. There are sites for this (www.clinicaltrials.gov). Unfortunately, I saw a paper recently which found that almost half of registered trials had significant differences between the registered and published methodologies, and no one was checking on this very carefully.


Flaws in one do not in any way strengthen the other, and for alternative medicine the evidence and studies are almost always either even weaker or non-existent.

exp(This).




Applications are open for YC Winter 2022

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: