The orthogonal projection of Ck onto the e?-space, reads
Jip dud, 0 1
u. 3% u; €
i
eO. e E ufe e.c
k jk CU (3.1.1)
p
k
then the projected vector C in the eP-space can be written as
Similarly the orthogonal projection of the observation vector Ax onto
the same subspace is
Lupi Jj i'. qup 19 i^ P Av XD
e (Ax =u; ee £34 e u: g Bii Ax. -ul Ax = Ay. (3.1.2)
and the projected vector can be written as
Tl p
Ay =
yo 7 9 Ej.
Figure 2 depicts these orthogonal projections
Fig. 2 The orthogonal projections of Ax and e onto e? -space :
The projected vectors e may not be linearly independent, it can be
proved that the following theorem holds /4/.
The e vectors are linearly independent in the condition space, i.e. eP -
space, if and only if the vectors Cr? So form a linearly independent set
in the observation space, i.e. e!-space.
Let us assume that the Cy are linearly independent, then they can be
considered as a covariant base of an m—dimensional subspace in the e? -
space. Hence, the e? -space may be split into two subspaces perpendicular
to each other, one, spanned by the vectors Cy and the other spanned by
(b-m) vectors =* perpendicular to the ee (see fig 2).
The covariant metric tensor of the Cr subspace is
- 99 -