Dear respected reader,
Perhaps you can help me on this one. Let X denote an n x p matrix which rank
is less than p. Let y denote an vector, and suppose I wish to solve the
equation y = Xb. A straightforward approach is setting
b = (X+)X'y,
where X+ denotes the Moore-Penrose inverse of X. My query concerns the
(asymptotic) dispersionmatrix of the coefficients b. Are these given by
cov(b) = (X+)V(X+),
where V denotes the residual dispersion cov(y|Xb)? Further, what would be a
suitable choice of V? Any literature and/or comments on the subject are
wellcome.
Thanks in advance,
Jarl Kampen
CESMO (Centrum voor Sociaal Methodologisch Onderzoek), Faculty of Social &
Political Sciences, Katholic University of Brussels, Vrijheidslaan 17, 1081
Brussels, Belgium. Voice +32(0)2 412 43 38. Fax +32(0)2 412 42 00. E-mail
[log in to unmask]
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|