文档介绍：c16v, 12/24/94
CHAPTER 16
OR THODO X METHODS: HISTORICAL BA CK GR OUND
\With al l this c onfounde dtr acking in hyp otheses ab out invisible c onne ctions with al l man-
ner of inc onc eivable pr op erties, which have che cke dpr o gr ess for so many ye ars, I b elieve
it to b e most imp ortant to op en p e ople's eyes to the numb er of sup er
uous hyp otheses they
ar e making, and would r ather exagger ate the opp osite view, if ne e db e, than pr o c e e d along
these false lines ." | on Helmholtz (1868).
This Chapter and Chapter 13 are concerned with the history of the sub ject rather than its presen t
status. There is plex and fascinating history b efore 1900, recoun ted b y Stigler (1986), but
w e are concerned no w with more recen t dev elopmen ts. In the p erio d from ab out 1900 to 1970,
one sc ho ol of though t dominated the eld pletely that it e to b e called \ortho do x
statistics". It is necessary for us to understand it, b ecause it is what most w orking statisticians
activ etoda yw ere taugh t, and its ideas are still b eing taugh t, and adv o cated vigorously , in man y
textb o oks and Univ ersities.
In Chapter 17 w ew an t to examine the \ortho do x" statistical practice th us dev elop ed and
compare its tec hnical p erformance with that of the \probabilit y as logic" approac h exp ounded
here. But rst, to understand this w eird course of ev en ts w e need to kno w something ab out the
problems faced then, the so ciology that ev olv ed to deal with them, the roles and p ersonalities of
the principal gures, and the general attitude to w ard scien ti c inference that ortho do xy represen ts.
The Early Problems
As w e note rep eatedly , the b eginnings of scien ti c inference w ere laid in the 18'th and 19'th Cen-
turies out of the needs of astronom y and geo desy . The principal gures w ere Daniel Bernoulli,
Laplace, Gauss, Legendre, P oisson and others, whom w ew ould describ e to da y as mathemati