I hope you will share the citations, links, or copies you receive.

My own crude approach includes trying to make sure that hypotheses 
have reasonably solid theoretical foundations and that the specific 
elements are as relevant to each other as possible.  I say "crude" 
because I mean that I favor rough, simple guesses rather than wild guesses.

Descriptive statistics are invaluable.  They can be better than rough 
guesses (hypothesis testing?) provided they are more relevant and 
sufficient to come to a conclusion, null or otherwise.

It seems to me that the question is not so much "either-or" but 
"which or both?"  It also seems that if a simple but honest test of 
an hypothesis short of descriptive statistics is sufficient or 
adequate to the task at hand, what, specifically is wrong with it?

Descriptive statistics have not always proved sufficient, for 
example, in the case of evaluating new drugs.  Application of 
theories, even conclusions reached by the best science can muster, 
can be flawed.  Actual experience seems to be, if not the final test, 
at least an "acid" one.  They have their utility, but, as the saying 
goes, "garbage in, garbage out."

But maybe statistics don't have to contain real garbage.  Perhaps all 
that is needed is for the relevant part to be in the outliers 
discarded by the program.  Flawed though it is, that flesh and blood 
computer ain't all bad.

WT


At 04:23 PM 2/26/2006, Malcolm McCallum wrote:
>I vaguely recall an exchange on here regarding the role of hypothesis =
>testing and the statistical validity of this approach.  Any citations or =
>comments would be greatly apprecieated!
>=20
>Malcolm L. McCallum
>Assistant Professor
>Department of Biological Sciences
>Texas A&M University Texarkana
>2600 Robison Rd.
>Texarkana, TX 75501
>O: 1-903-233-3134
>H: 1-903-791-3843
>Homepage: https://www.eagle.tamut.edu/faculty/mmccallum/index.html
>=20

Reply via email to