Understanding statistical analysis in PET paper

NMF nm_fournier at ns.sympatico.ca
Tue Mar 2 02:19:38 EST 2004

> That is, in condition 1, activation at locus A de-
> creases, but activation at locus B can go up or
> down, and in condition 2, activation at locus A
> increases, but activation at locus B can go up
> or down.
> So nothing can be said about locus A, without
> knowing what goes on at =every= "locus B".

Granted.  I agree with you.  Guess what it is no surprise why you would come
to these conclusions.  More often than not many scientists do not have a
firm grasp of statistical theory.  Many neuroscientists assess multivariate
problems using univariate approaches.  It's quite sickening.  Bivariate
correlations between a single brain structure and a behavior are a silly
methodological approach to discern brain behavioral correlates.  Behavior is
multivariate process involving the immense complexity of parallel circuitry
proliferated across a variety of multifocal aggregates that are embedded
within brain space. Statistical approaches must take into account this
aspect of organization (and the functional dynamics, you have just

However, you are wrong about one thing.  There are MANY statistical
techniques and approaches that you can do to assess these exact points that
are you raising.  Partial correlations, canonical correlations,  and
multiple regression would be one approach.  Also the use of lag-lead
analyses with autocorrelations could be employed with non-linear statistical
regression methods, is another.  I think you have greatly underestimated the
over thousand years of mathematics at our disposal.

Based upon the approach you have just described your model has one built-in
limitation.  It may theorize regarding how the nervous system all it likes,
but it cannot and never provide any significant prediction regarding what
the nervous system is doing at any given moment.  In other words, you
cannot -based upon the premise you have suggested above- predict any aspect
of neuronal dynamics.  You may provide an insight into the potential
principles of neuronal functioning but you can never provide any real
information regarding how the nervous system works. The processes you
suggest would operate, however, you can never say anything more than that.
You can never predict behavior based upon your theoretical framework.
Prediction is a quantitative process, something that could never be insured
with any absolute based upon the theory that you have presented.  (It's
almost quantum mechanical in some aspects).

I'm sorry, but I disagree with your logic here. You did most of research in
the 1970's, right?   A lot of things have happened and changed regarding how
people think and evaluate neuronal phenomena.

More information about the Neur-sci mailing list