Science

[Frederick G Scoles]

On Thu, 27 Jul 1995 rhohf@eagle.idcnet.com wrote>
>
>
> There is another requirement of the "scientific method" which is apparently
> not being met:  the >isolation of variables<.  This means, "Measure one thing
> while everything else remains unchanged," and is, perhaps, the most
> fundamental principle of the scientific method. Some reflection on this point
> will indicate that science presents a very skewed view of the universe; there
> are many things (perhaps including piano actions) which simply have too many
> variables to isolate and measure them. Thus, the scientific method is more
> useful for some things than for others.  Attempts to employ the scientific
> method improperly result in descriptions of >special cases< with little or no
> general application.  It may be that every piano ever made is its own special
> case. Solving action problems to the extent of producing consistent and
> predictable results may require seeing through the mounds of data to a deeper
> level of understanding.
>
> Bob Hohf
>
I agree with the above, but modern experimental methods have modified
this since about 1932 when Sir Ronald Fisher of Rothamsted Expt. Stn.
published his early  one-way ANOVA  for agricultural work.  It usually
costs far too much time and money to vary only one factor at a time.
Instead, it's better to design the study using an appropriate
experimental (statistical) design, so that one can vary quite a few
factors simutaneously and allow the statistics to separate everything.
Each variable will have its own standard deviation and sum of squares.
Many basic designs exist with ANOVA's (Analysis of Variance), factorials,
regressions, etc. which are free on many software packages.(Excel, SAS,
etc.)  It is true
that sciences require the separation of each factor, but this is only a
statistical requirement.  Therefore, many things can be varied
simultaneously as long as the stats. can handle it.  The bottom line is
usually how much time and money do you have.  Also, most colleges and
universities have good statisticians who can tell us which design to use,
and most of them enjoy helping us out; especially ones like me who took
the research statistics but have forgotten it all. Statistics designs are
there to help save grief, time, and money.

Fred Scoles,  RPT