Jump to content

  • Log in with Facebook      Sign In   
  • Create Account       


Why almost 40% of peer-reviewed dietary research turns out to be wrong

statistics significance hidden variables

  • Please log in to reply
2 replies to this topic

#1 ras

  • Supporting
  • 11 posts

Posted 21 May 2018 - 06:31 PM

Good non-technical article on what's wrong with p=.05:



-Richard Schulman
 Editor, https://foundsbroadsheet.com

#2 TomBAvoider

  • Member
  • 546 posts

Posted 22 May 2018 - 01:37 AM

Hard to argue with, sadly. The whole field is a mess. But I'm not optimistic there's enough will to clean it up.

#3 mccoy

  • Member
  • 1,183 posts

Posted 23 May 2018 - 04:46 AM

I'm learning to be very skeptical with so called scientific articles and, beyond the screening methods suggested above, I also use intuitive degree of beliefs and a priori bayesian considerations.

Dempster-Shafer theory

Bayesian decision theory


In a few words, if an article comes up illustrating the detrimental effects of fresh fruit, based on my subjective a priori knowledge and on the existing degrees of beliefs in holistic medicine, I'll label it as probably hogwash.


I may also want to experiment on it, that is add data or evidence, or Bayesian likelyhood. For example, I experimented on myself that 600 grams of oranges do not cause a glycaemic peak. So what may be all that detrimental effect? The Lustig hypothesis of detrimental liver metabolism of fructose? I get back to prior experience: I'm thriving and my liver is presently perfectly healthy after 40 years of abundant ingestion of fresh fruit.

I get back to colelcting more evidence, and it seems that orange (and fruit) has not only fructose, but significant glucose and sucrose and fiber of course. So it's a far cry from pure fructose, which may actually be detrimental. And articles have come up that comment about the possible detrimental effects of fructose conditioned to the fact that the energy requirements are significantly surpassed. Not my case. Not everyone's case.


The above is an example of how I intuitively and practically tackle the literature issues, it may not be done for all topics, but I never accept so called scientific evidence as truth. Simply because that's hardly undisputable scientific evidence. That's the results of a particular study or experiment, which may be biased, may be statistically not enough representative, may be difficult to intepret, may have hidden or manifest agendas and so on. Not an easy issue to figure out, but skepticism remains the password.

Edited by mccoy, 23 May 2018 - 04:49 AM.

"Data speak for themselves" -Reverend Thomas Bayes 1702-1761

Also tagged with one or more of these keywords: statistics, significance, hidden variables