文档介绍：y
Quan tile Regression
Roger Ko enk er
Departmen t of Economics
Univ ersit y of Illinois, Urbana-Champaign
Champaign, IL 61820 USA
V ersion: No v em b er 10, 2000. This material constitutes v ery preliminary v ersions of sev eral
c hapters from a prop osed monograph on quan tile regression. P ortions of the presen tv ersion w ere
prepared for a minicourse o ered b y Roger Ko enk er in F ebruary 1997 at Camp os dos Jord~ ao, Sao
P aulo, Brazil. Commen ts on an y asp ect of the c hapters w ould b e most w e and could b e sen t
to ******@ or to the p ostal address pro vided ab o v e.
1
2
Con ten ts
3
CONTENTS i
Preface
F rancis Galton in a famous passage defending the \c harms of statistics" against
its man y detractors, c hided his statistical colleagues
[who] limited their inquiries to Av erages, and do not seem to rev el in more
comprehensiv e views. Their souls seem as dull to the c harm of v ariet y
as that of a nativ e of one of our
at English coun ties, whose retrosp ect
of Switzerland w as that, if the moun tains could b e thro wn in to its lak es,
t w on uisances w ould b e got rid of at once. [ Natur al Inheritanc e, p. 62]
It is the fundamen tal task of statistics to bring order out of the div ersit y , at times
the apparen tc haos, of scien ti c observ ation. And this task is often v ery e ectiv ely
plished b y exploring ho w aver ages of certain v ariables dep end up on the v alues
of other \conditioning" v ariables. The metho d of least squares whic hperv ades sta-
tistics is admirably suited for this purp ose. And y et, lik e Galton, one ma y question
whether the exclusiv e fo cus on conditional mean relations among v ariables, ignores
some \c harm of v ariet y" in matters statistical.
As residen ts of one of the
attest American coun ties, our recollections of Switzer-
land and its attractiv en uisances are quite di eren t from the ones describ ed b y Galton.
Not only the Swiss landscap e, but