Posted by Larry Hoover on May 15, 2006, at 22:37:15
In reply to Re: Statistical question on SSRIs - ADDENDUM » Larry Hoover, posted by Squiggles on May 15, 2006, at 16:57:28
> ...........
>
>
> > Statistical significance is not proof of anything at all. One underlying assumption that never goes to zero is that the results are purely and entirely due to chance.
> >
> > If you sample a population enough times, you can always find a significant result, no matter how absurd the hypothesis being tested.
> >
> > Much of the research that has been published is not proof of anything at all.
> >
> > Lar
>
>
>
> Is this an opinion that you hold Larry or
> is it one that the medical community and
> especially the psychopharmacology community
> goes by?I was speaking my opinion, of course. How could I speak for these other people I've never met? I'm talking about critical thinking. About the limits of scientific "proof". People like the idea that there might be proof, but science doesn't very often prove anything at all. When we say something is statistically significant, the very statement itself includes an assumption that the difference found between groups could be chance. That's what p < .05 indicates. Or, p < .01, or whatever. p is (virtually) never zero.
> Because Dr. Nemeroff (and the others,
> e.g. Kessler) in this early 1991 FDA hearing on suicidality and antidepressants, states the following:
>
> "The real issue is how can we, scientifically, as a profession, come to grips with this difficult issue? Clearly, what we need are double-blind placebo-controlled trials. I would like to read a quote from David Kessler, Commissioner Od
> the FDA. In his recent article in the New England Journal of Medicine, he said, "scientific rigor requires that data presented during an activity be reliable, that is, capable of
> forming an appropriate basis for medical decision making."You can still make a scientific decision, absent proof. Scientists know that they aren't basing decisions on proof, the way lay people know proof. This isn't fingerprint analysis, or DNA analysis we're discussing. (And even those have error estimates attached.)
> Scientifically rigorous data are developed through study designs that minimize bias. Anecdotal evidence and unsupported opinion should play no part in a scientifically rigorous program."
"Anecdote.....should play no part..."
Precisely.
> So the issue, then, is what, on the one hand, can we learn from case reports and anecdotal data, and I think it car give us a signal for prospective studies. I remind all of you that the history of medicine is replete with examples of
> medical decision making based on anecdotal case reports, to wit, the use of widespread tonsillectomies in all of our
> children -- at least not our children, but there are very few people in this room who have tonsils. We have now discovered that that was unnecessary surgery, and how did we discover it? By prospective controlled trials.".....
>
> pp. 201-202For years, tonsils were routinely removed because a child had a sore throat. Why? Because that's what doctors thought they ought to do. Finally, some courageous souls stood up for scientific rigour, and it was shown that tonsillectomy did not improve a child's health. I.e. p *greater than* 0.05. Some children do benefit from the surgery, and it is still performed. It is no longer routine, however.
> I don't mean to say that I know what scientific
> proof is, but it certainly seems that statistical significance is the lingua franca of what is acceptable as scientifically valid today.You have to consider what is being shown to be significant, also. Far too often, unproven conclusions are published, as if they had been proven. A significant outcome was demonstrated, but it didn't support that which was concluded.
> Even if this method proved nothing, it certainly has a great impact on public health care if it is universally accepted by clinicians and doctors.
>
> SquigglesI have been struggling, as I read your post, to find your argument. I didn't find one. Could you please try again?
Lar
poster:Larry Hoover
thread:640557
URL: http://www.dr-bob.org/babble/20060515/msgs/644539.html