Re: Who "invented" the null hypthesis?


"Dieter Folz" <dieterfolz@xxxxxx> wrote in message
news:1146206361.762951.35150@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
Hi everyone!

Talking about stats, some time during the explanations I was
practically asked by a friend what the origins and the fundamental
logic of the null hypothesis was. I didn*t really know the answer. In
fact I started to sweat as I explainied that you want to prove that,
let's say two groups differ significantly in a measure (body weight for
example) from each other (=your hypothesis, called H_1). Therefore,
due to aristotelian logic(?), (hegelian?) dialectic(?) or poppersian
falsificationism(?), you choose a null hypothsis (H_0) which says the
contratictory thing to H_1 (here: no difference). Then your finding
(which of course should prove your real hypothesis H_1) must be THAT
unlikely under the assumption that H_0 is true, that we have to say,
considerig an eror of x% (mostly 5%), the H_0 can't be true, because it
then is so immensly unlikely that we would have found these kind of
effects.

Apart from the allegation of cheating to just choose an H_0 to
pretend to do a falsification, I had more and more problems to explain
the concept clearly, and had more and more the impression that I don't
understand it myself correctly nor sufficient. Esp. I have no idea what
the real reason or basic idea of the null hypothesis is, what the
exact, specific and datailed ideas, mathematical and logical
backgrounds are.

Who "invented" the null hypothesis, whre does it come from, which
concepts are the basis for it? Hope, somebody can help me out.


Cheers, Dieter
++++++++++++++++++++++++++++++++++++++++++++++
The book "What If There Were No Significance Tests?" by Harlow, Mulaik and
Steiger (Erlbaum 1997) is recommended for backgraound on the essential
problems in its use today..

Harlow points out that Karl Pearson in 1901 laid the groundwork for a
scientific hypothesis with sample data. Fisher then proposed a set of
methods for testing (ca 1930) based on p values. Then of course Neyman came
out with the basic method (1928). The book "The Lady Tasting Tea", describes
the uproar in the Royal Society when Neyman proposed his theory and was
challenged. The controversy between Fisher and Neyman in the 1930's is part
of the interesting history here.

Hubbard and Bayarri's TAS article ("Confusion Over Measures of Evidence
(p's) Versus Error (alpha) in Classical Statistical Testing") in the Aug
2003 issue of TAS describes the differences between the Fisher view,
Neyman's views, and the horrible mixup of ideas that is the current
hypothesis test. Serlin's article in JASM, vol 1, no. 2 puts it all in the
historic view of the pusuit of a philosophy of science.

Neyman's original view was that hyposthesis testing was only to be used a a
tool for quality control in manufacturing processes. His terminology and
usage can only be properly understood from this context.

DAH


.

Loading