The least-square estimate of is found by minimizing Q. Applying the
variational principle discussed in Section 9.3.1.1, we have
Setting the solution for is obtained from normal equation
or
which gives
In the above, the inverse of matrix CTC exists if there are at least two distinct
values of xi represented in the sample.
We can easily check that Equation (11.15) is identical to Equations (11.7)
and (11.8) by noting that
and
340 Fundamentals of Probability and Statistics for Engineers
q,^q,
dQdqTCT
yCq
yCqTCdq
2 dqTCT
yCq:
dQ0, ^q
CT
yC^q 0 ;
11 : 14
CTC^qCTy;
^q
CTC^1 CTy:
11 : 15
CTC
11
1
x 1 x 2 xn
1 x 1
1 x 2
... ...
1 xn
2 6 6 6 6 4
3 7 7 7 7 5
nnx
nx
Pn
i 1
x^2 i
2
4
3
(^5) ;
CTy
11
1
x 1 x 2 xn
y 1
y 2
...
yn
2 6 6 6 6 4
3 7 7 7 7 5
ny
Pn
i 1
xiyi
2
4
3
(^5) ;
^q
CTC^1 CTy
nnx
nx
Pn
i 1
x^2 i
2
4
3
5
(^1) ny
Pn
i 1
xiyi
2
4
3
5
y
Pn
i 1
xiyinxy
Pn
i 1
x^2 inx^2
1
x
Pn
i 1
xiyinxy
Pn
i 1
x^2 inx^2
1
2
(^66)
(^64)
3
(^77)
(^75)
y ^x
Pn
i 1
xix
yiy
Pn
i 1
xix^2
1
2
(^64)
3
(^75) :