Kjetil,
I found the following historical information about
the Gauss-Markov Theorem on the web:
---
The name GAUSS-MARKOV THEOREM for the
chief result on least squares and best linear unbiassed estimation in the linear
(regression) model has a curious history. David (1998) refers to H. Scheff�'s
1959 book Analysis of Variance for the first use of the phrase "Gauss-Markoff
theorem" although a JSTOR search finds a few earlier occurrences including one
from 1951 by E. L. Lehmann (Annals of Mathematical Statistics, 22, No. 4, p.
587). For some years previously the term "Markoff theorem" had been in use. It
was popularised by J. Neyman who believed that this Russian contribution had
been overlooked in the West--see his "On the Two Different Aspects of the
Representative Method" (Journal of the Royal Statistical Society, 97, (1934),
558-625). The theorem is in chapter 7 of the book translated into German
as Wahrscheinlichkeitsrechnung (1912). However R. L.
Plackett (Biometrika, 36, (1949), 149-157) pointed out that Markov had done no
more than Gauss nearly a century before in his Theoria combinationis observationum erroribus minimis
obnoxiae (1821/3). (In the nineteenth century the
theorem was often referred to as "Gauss's second proof of the method of least
squares"--the "first" being a Bayesian argument Gauss published in 1809 based on
the normal distribution). Following Plackett, a few authors adopted the
_expression_ "Gauss theorem" but "Markov" was well-entrenched and the compromise
"Gauss-Markov theorem" has become standard.
This entry was contributed by John
Aldrich.
---
----- Original Message -----
From: <[EMAIL PROTECTED]>
To: <[EMAIL PROTECTED]>
Sent: Wednesday, May 05, 2004 1:17 PM
Subject: [edstat] Gauss-Markov
theorem
> but when/where did Markov get into it?
> And, is it the same Markov as in "Markov chains"?
>
> Kjetil Halvorsen
> .
> .
> =================================================================
> Instructions for joining and leaving this list, remarks about the
> problem of INAPPROPRIATE MESSAGES, and archives are available at:
> . http://jse.stat.ncsu.edu/ .
> =================================================================
>
