Posted by Larry Hoover on December 16, 2005, at 9:41:26
In reply to Re: What!, posted by linkadge on December 15, 2005, at 19:15:42
> Correlation does not imply causation.
>
> Suicide rates go up and down.
>
> LinkadgeHmmm. It's hard to guess just what it was in my lengthier post that was the progenator of this brief response.
Tracy says SSRIs kill, via suicide. SSRI presciption rates have skyrocketed. This implies?
We look at available data, independently collected by reputable scientists for reputable motives, and we do not see the implied signal.
To the stats. Correlation is not evidence of causation. I would argue that it does imply it, based on my perception of human nature.
However, correlation has four categories of explanation.
1. Some variable A leads to some measurable change in variable B.
2. Some variable B leads to some measurable change in variable A.
3. It was a fluke. There is no relationship between A and B, but you measured a group unrepresantative of the greater population.
4. Some unmeasured variable C, leads to measurable change in both variable A and variable B.This is where the tough thinking begins. Hypotheses form. The thing is, the first 2 cases are indistinguishable. It doesn't matter how you graph them (which variable is the arbitrary independent variable, on axis X, and which is the dependent, graphed on Y), the very same coefficient of correlation will be obtained, with indentical significance.
What most people also fail to consider, even when numerous experiments have narrowed it down to e.g. A probably causes B, you can never exclude the fourth explanation.
Scientists do, however, love to assign probability to their data, i.e. the probability that their data is not a fluke observation.
It is my opinion that the probability that there has been a net suicide increase due to SSRI exposure is close to nil. Unfortunately, tragedies have occurred. And I will always be saddened by that realization. I think we have let people down. I pray we bring all our resources together to prevent all that we can. By looking at all available data, we now know when the vulnerable periods are.
Lar
poster:Larry Hoover
thread:587690
URL: http://www.dr-bob.org/babble/20051211/msgs/589551.html