Department of Geospatial Science RMIT

2. THE GENERAL LEAST SQUARES ADJUSTMENT TECHNIQUE

A common treatment of the least squares technique of estimation starts with simple linear
mathematical models having observations (or measurements) as explicit functions of parameters
with non-linear models developed as extensions. This adjustment technique is generally described
as adjustment of indirect observations (also called parametric least squares). Cases where the
mathematical models contain only measurements are usually treated separately and this technique is
often described as adjustment of observations only. Both techniques are of course particular cases
of a general adjustment model, the solution of which is set out below. The general adjustment
technique also assumes that the parameters, if any, can be treated as "observables" ie, they have an
a priori covariance matrix. This concept allows the general technique to be adapted to sequential
processing of data where parameters are updated by the addition of new observations.

In general, least squares solutions require iteration, since a non-linear model is assumed. The
iterative process is explained below. In addition, a proper treatment of covariance propagation is
presented and cofactor matrices given for all the computed and derived quantities in the adjustment
process. Finally, the particular cases of the general least squares technique are described.

2.1. The General Least Squares Adjustment Model

Consider the following set of non-linear equations representing the mathematical model in an
adjustment

d i
F l, x = 0 (2.1)

where l is a vector of n observations and x is a vector of u parameters; l and x referring to

estimates derived from the least squares process such that

l = l + v (2.2)

x = x + δ x (2.3)

where v is a vector of residuals or small corrections and δx is a vector of small corrections.

© 2003, R.E. Deakin Coordinate Transformations (2003) 2–1

Department of Geospatial Science RMIT

As is usual, the independent observations l have an a priori diagonal cofactor matrix Qll containing
estimates of the variances of the observations, and in this general adjustment, the parameters x are
treated as "observables" with a full a priori cofactor matrix Q xx . The diagonal elements of Q xx
contain estimates of variances of the parameters and the off-diagonal elements contain estimates of
the covariances between parameters. Cofactor matrices Qll and Q xx are related to the covariance
matrices Σ ll and Σ xx by the variance factor σ 20

Σ ll = σ 20 Qll (2.4)

Σ xx = σ 20 Q xx (2.5)

Also, weight matrices W are useful and are defined, in general, as the inverse of the cofactor
matrices

W = Q −1 (2.6)

and covariance, cofactor and weight matrices are all symmetric, hence Q T = Q and W T = W
where the superscript T denotes the transpose of the matrix.

Note also, that in this development where Q and W are written without subscripts they refer to the
observations, i.e., Qll = Q and Wll = W
Linearizing (2.1) using Taylor's theorem and ignoring 2nd and higher order terms, gives

d i
F l, x a f
= F l, x +
∂F
∂l l, x
dl − li + ∂∂Fx bx − xg
l, x
= 0 (2.7)

and with v = l − l and δx = x − x from (2.2) and (2.3), we may write the linearized model in
symbolic form as

Av + B δ x = f (2.8)

Equation (2.8) represents a system of m equations that will be used to estimate the u parameters
from n observations. It is assumed that this is a redundant system where

n ≥ m ≥ u (2.9)

© 2003, R.E. Deakin Coordinate Transformations (2003) 2–2

Department of Geospatial Science RMIT

The redundancy or degrees of freedom is

r = m−u (2.10)

In equation (2.8) the coefficient matrices A and B are design matrices containing partial derivatives
of the function evaluated using the observations l and the "observed" parameters x.

∂F
A m,n = (2.11)
∂l l, x

∂F
Bm , u = (2.12)
∂x l, x

The vector f contains m numeric terms calculated from the functional model using l and x.

m a fr
fm,1 = − F l, x (2.13)

2.2. The Least Squares Solution of the General Adjustment Model

The least squares solution of (2.8), ie, the solution which makes the sums of the squares of the
weighted residuals a minimum, is obtained by minimising the scalar function ϕ

a
ϕ = v T W v + δ x T Wxxδ x − 2kT Av + Bδ x − f f (2.14)

where k is a vector of m Lagrange multipliers. ϕ is a minimum when its derivatives with respect to
v and δx are equated to zero, ie.

∂ϕ
= 2v T W − 2kT A = 0T
∂v
∂ϕ
= 2δ x T Wxx − 2k T B = 0 T
∂δx

These equations can be simplified by dividing both sides by two, transposing and changing signs to
give

− Wv + A T k = 0 (2.15)

− Wxxδ x + BT k = 0 (2.16)

© 2003, R.E. Deakin Coordinate Transformations (2003) 2–3

Department of Geospatial Science RMIT

Equations (2.15) and (2.16) can be combined with (2.8) and arranged in matrix form as

LM−W AT 0 OP L v O L0O
MM A 0 B PP MM k PP = MMf PP (2.17)
N0 BT − Wxx Q MNδ xPQ MN0PQ
Equation (2.17) can be solved by the following reduction process given by Cross (1992, pp. 22-23).
Consider the partitioned matrix equation P y = u given as

LMP 11 P12 OP LM y OP = LM u OP
1 1
(2.18)
NP 21 P22 Q N y Q Nu Q
2 2

which can be expanded to give

P11 y1 + P12 y 2 = u1

or

b
y1 = P11−1 u1 − P12 y 2 g (2.19)

Eliminating y1 by substituting (2.19) into (2.18) gives

LMP
11 P12 OP LMP bu − P y gOP = LM u OP
−1
11 1 12 2 1

NP
21 P22 QN y Q Nu Q 2 2

Expanding the matrix equation gives

b
P21P11−1 u1 − P12 y 2 + P22 y 2 = u 2 g
P21P11−1u1 − P21P11−1P12 y 2 + P22 y 2 = u 2

and an expression for y 2 is given by

cP22 h
− P21P11−1P12 y 2 = u2 − P21P11−1u1 (2.20)

Now partitioning (2.17) in the same way as (2.18)

LM−W AT 0 OP L v O L0O
MM A 0 B PP MM k PP = MMf PP (2.21)
N0 BT − Wxx Q MNδ xPQ MN0PQ
v can be eliminated by applying (2.20)

LML 0 B OP LM OP
A W −1 A T 0 OP L k O = Lf O − LAO −W −1
0
NMNB Q MNδ xPQ MN0PQ MN 0 PQ
−
T
− Wxx Q NQ
0

© 2003, R.E. Deakin Coordinate Transformations (2003) 2–4

Department of Geospatial Science RMIT

Remembering that Q = W −1 the equation can be simplified as

LMAQA T
B OP L k O = Lf O
Q MNδ xPQ MN0PQ
(2.22)
NB T
− Wxx

Again, applying (2.20) to the partitioned equation (2.22) gives

e−W xx c
− BT AQA T h Bj δ x = 0 − B cAQA h
−1 T T −1
f

and re-arranging gives the normal equations

eB cAQA h
T T −1
j
B + Wxx δ x = BT AQA T c h −1
f (2.23)

Mikhail (1976, p. 114) simplifies (2.23) by introducing equivalent observations le where

le = A l (2.24)

Applying the matrix rule for cofactor propagation (Mikhail 1976, pp. 76-79) gives the cofactor
matrix of the equivalent observations as

Q e = AQA T (2.25)

With the usual relationship between weight matrices and cofactor matrices, see (2.6), we may write

We = Q e−1 = AQA T c h −1
(2.26)

Using (2.26) in (2.23) gives the normal equations as

cB W B + W h δ x = B W f
T
e xx
T
e (2.27)

With the auxiliaries N and t

N = BT We B (2.28)

t = BT We f (2.29)

the vector of corrections δ x is given by

δ x = N + Wxx b g −1
t (2.30)

© 2003, R.E. Deakin Coordinate Transformations (2003) 2–5

Department of Geospatial Science RMIT

The vector of Lagrange multipliers k are obtained from (2.22) by applying (2.19) to give

c
k = AQA T h af − B δ x f = W af − B δ x f
−1
e (2.31)

and the vector of residuals v is obtained from (2.21) as

− Wv + A T k = 0

giving

v = W −1A T k = QA T k (2.32)

2.3. The Iterative Process of Solution

Remembering that x = x + δ x , see (2.3), where x is the vector of a priori estimates of the
parameters, δ x is a vector of corrections and x is the least squares estimate of the parameters.

At the beginning of the iterative solution, it can be assumed that x equals the a priori estimates x1
and a set of corrections δ x1 computed. These are added to x1 to give an updated set x 2 . A and B
are recalculated and a new weight matrix Wxx computed by cofactor propagation. The corrections
are computed again, and the whole process cycles through until the corrections reach some
predetermined value, which terminates the process.

x n +1 = x n + δ x n (2.33)

2.4. Derivation of Cofactor Matrices

In this section, the cofactor matrices of the vectors x , δ x, v and l will be derived. The law of
cofactor propagation will be used and is defined as follows (Mikhail 1976, pp. 76-89).

Given a functional relationship

z= Fx af (2.34)

between two random vectors z and x and the variance-covariance matrix Σ xx , the variance-
covariance matrix of z is given by

Σ zz = J zx Σ xx J Tzx (2.35)

© 2003, R.E. Deakin Coordinate Transformations (2003) 2–6

Department of Geospatial Science RMIT

where J zx is a matrix of partial derivatives

LM ∂z 1 ∂z1
"
∂z1 OP
MM ∂∂zx 1 ∂x2
∂z2
∂xn
∂z2
PP
∂F
= M ∂x PP
2
"
J zx = ∂x2 ∂xn
∂x M 1

MM ∂z# m ∂zm
#
∂zm PP
"
MN ∂x 1 ∂x2 ∂xn PQ
Using the relationship between variance-covariance matrices and cofactor matrices, see (2.5), the
law of cofactor propagation may be obtained from (2.35), as

Q zz = J zx Q xx J Tzx (2.36)

For a function z containing two independent random variables x and y with cofactor matrices Q xx
and Q yy

z = F x, ya f (2.37)

the law of cofactor propagation gives the cofactor matrix Q zz as

Q zz = J zx Q xx J Tzx + J zy Q yy J Tzy (2.38)


2.4.1. Cofactor Matrix for x

According to equations (2.33) and (2.30) with (2.29) the least squares estimate x is

b
x = x + N + Wxx g −1
BT We f (2.39)

and x is a function of the a priori parameters x (the "observables") and the observations l since the
vector of numeric terms f contains functions of both. Applying the law of propagation of cofactors
gives

F ∂x I F ∂x I F ∂x I F ∂x I

= G J Q G J + G J QG J
T T

Q xx 
H ∂x K H ∂x K H ∂l K H ∂l K
xx (2.40)

The partial derivatives of (2.39) are

∂x ∂f
∂x
b
= I + N + Wxx g −1
BT We
∂x
(2.41)

© 2003, R.E. Deakin Coordinate Transformations (2003) 2–7

Department of Geospatial Science RMIT

∂x ∂f
∂l
b
= N + Wxx g −1
BT We
∂l
(2.42)

a f
From (2.13), f = − F x, l the partial derivatives
∂f
∂x
∂f
and , are the design matrices A and B given
∂l
by (2.11) and (2.12)

∂f
= −B (2.43)
∂x

∂f
= −A (2.44)
∂l

Substituting (2.43) and (2.44) into (2.41) and (2.42) with the auxiliary N = BT We B gives

∂x
∂x
b
= I − N + Wxx g −1
BT We B
(2.45)
= I − bN + W g
−1
xx N

∂x
∂l
= − N + Wxxb g −1
BT We A (2.46)

Substituting (2.45) and (2.46) into (2.40) gives

o b g Nt Q oI − bN + W g Nt
−1 −1 T
Q xx  = I − N + Wxx xx xx
(2.47)
+ o−bN + W g B W At Q o−bN + W g B W At
−1 −1 T
T T
xx e xx e

With the auxiliary

•
N = N + Wxx b g (2.48)
•
and noting that the matrices I, N, N and Wxx are all symmetric, (2.47) may be simplified as

FG • IJ FG • IJ FG • IJ FG • IJ
H K H K H
Q xx  = I − N −1 N Q xx I − N N −1 + N −1 BT We A Q A T We B N −1
K H K
Remembering that Q e = AQA T and We = Q e−1

• • • • • •
Q xx  = Q xx − Q xx N N −1 − N −1 NQ xx + N −1 NQ xx N N −1 + N −1 N N −1 (2.49)

The last two terms of (2.49) can be simplified as follows

© 2003, R.E. Deakin Coordinate Transformations (2003) 2–8

Department of Geospatial Science RMIT

• • • • • FG IJ• •
N −1 NQ xx N N −1 + N −1 N N −1 = N −1 NQ xx N N −1 + Wxx N −1
H K
•
= N −1 NQ xx bN + W g N xx
• −1

• • •
= N −1 NQ xx N N −1
•
= N −1 NQ xx

and substituting this result into (2.49) gives

• • •
Q xx  = Q xx − Q xx N N −1 − N −1 NQ xx + N −1 NQ xx
(2.50)
• −1
= Q xx − Q xx N N

Further simplification gives

FG IJ •

H
Q xx  = Q xx I − N N −1
K
= Q xx FH N− NIK N
• • −1

(2.51)

= Q xx bN + W − Ng N xx
• −1

•
= Q xx Wxx N −1

and since Q xx Wxx = I the cofactor matrix of the least squares estimates x is

•
Q xx  = N −1 = N + Wxx b g −1
(2.52)

2.4.2. Cofactor Matrix for l

Beginning with the final adjusted observations given by (2.2)

l = l + v (2.53)

and using (2.32) and (2.31) we have

l = l + QA T k
= l + QA T We f − Bδ x a f
= l + QA We f − QA We Bδ x
T T

© 2003, R.E. Deakin Coordinate Transformations (2003) 2–9

Department of Geospatial Science RMIT

•
Substituting the expression for δ x given by (2.30) with the auxiliaries t and N given by (2.29) and
(2.48) respectively gives

e b e g
l = l + QA T W f − QA T W B N + W
xx
−1
t
= l + QA W f − QA W B bN + W g
T T −1
e e xx BT We f (2.54)
•
= l + QA T We f − QA T We B N −1 BT We f

and l is function of the observables x and the observations l since f = − F x, l . Applying the law a f
of propagation of cofactors to (2.54) gives

Qll
F ∂l I F ∂l I F ∂l I F ∂l I
= G J Q G J + G J QG J
T T

(2.55)
H ∂x K H ∂x K H ∂l K H ∂l K
xx

and the partial derivatives are obtained from (2.54) as

∂l ∂f • ∂f
= Q A T We − Q A T We B N −1 BT We
∂x ∂x ∂x

∂l ∂f • ∂f
= I + Q A T We − Q A T We B N −1 BT We
∂l ∂l ∂l

∂f ∂f
With = − B and = − A , and with the auxiliary N = BT We B the partial derivatives become
∂x ∂l

∂l •
= Q A T We B N −1 BT We − Q A T We B
∂x (2.56)
•
= Q A T We B N −1 N − Q A T We B

∂l •
= I + Q A T We B N −1 BT We A − Q A T We A (2.57)
∂l

Substituting (2.56) and (2.57) into (2.55) gives

m r m
Qll = 1st term + 2 nd term r (2.58)

where

© 2003, R.E. Deakin Coordinate Transformations (2003) 2–10

Department of Geospatial Science RMIT

m1st
r • •
term = QA T We B N −1 NQ xx N N −1 BT We AQ − QA T We B N −1 NQ xx BT We AQ
•

•
− QA T WeBQ xx N N −1 BT We AQ + QA T We BQ xx BT We AQ

m2 nd
r
term = Q + QA T We B N −1 BT We AQ − QA T We AQ
•

•
+ QA T We B N −1 BT We AQ
• •
+ QA T We B N −1 BT We AQA T We B N −1 BT We AQ
•
− QA T We B N −1 BT We AQA T We AQ − QA T We AQ
•
− QA T We AQA T We B N −1 BT We AQ + QA T We AQA T We AQ

The 1st term can be simplified as

m1 r FG • • • • IJ
H
term = QA T We B N −1 NQ xx N N −1 − N −1 NQ xx − Q xx N −1 N + Q xx BT We AQ
K
st

F F
= QA W B G N N G Q
• −1 • IJ − Q • IJ
H H N N −1 − Q xx
K N −1 N + Q xx BT We AQ
T
e xx xx
K
• •
but we know from (2.50) that Q xx  = Q xx − Q xx N N −1 , and from (2.52) that Q xx  = N −1 so

m1 r FG • IJ
H
term = QA T We B Q xx  − N −1 NQ xx  BT We AQ
K
st

F
= QA W B G N
• −1 • • IJ B W AQ
H − N −1 N N −1
K
T T
e e

• FG • IJ
= QA T We B N −1 I − N N −1 BT We AQ
H K
•
The term in brackets has been simplified in (2.51) as Wxx N −1 which gives the 1st term as

m1 st
r
term = QA T We B N −1 Wxx N −1 BT We AQ
• •
(2.59)

The 2nd term of (2.58) can be simplified by remembering that AQA T = Q e = We−1 so that after
some cancellation of terms we have

m2 nd
r •
term = Q + QA T We B N −1 N N −1 BT We AQ − QA T We AQ
•
(2.60)

Substituting (2.59) and (2.60) into (2.58) gives the cofactor matrix of the adjusted observations as

Qll = Q + QA T We B N + Wxx b g −1
BT We AQ − QA T We AQ (2.61)

© 2003, R.E. Deakin Coordinate Transformations (2003) 2–11

Department of Geospatial Science RMIT

2.4.3. Cofactor Matrix for δ x

From (2.30) and (2.29)

b
δ x = N + Wxx g −1
BT We f
(2.62)
• −1
= N B We fT

and applying the law of propagation of cofactors gives

FGIJ FG
• IJ •
T

H K H K
Qδ xδ x = N −1 BT We Q f f N −1 BT We (2.63)

The cofactor matrix Q f f is obtained from f = − Fa x, lf as

F ∂f I F ∂f I F ∂f I F ∂f I
Q = G J Q G J + G J QG J
T T

ff
H ∂x K H ∂x K H ∂l K H ∂l K
xx

= a− Bf Q a − Bf + a− Af Q a− Af
T T
xx (2.64)
= BQ xx BT + AQA T
= BQ xx BT + Q e

Substituting (2.64) into (2.63) and simplifying gives

b
Qδ xδ x = N + Wxx g −1
b
NQ xx N N + Wxx g + bN + W g N bN + W g
−1
xx
−1
xx
−1
(2.65)

Equation (2.65) can be simplified further as

• • • •
Qδ xδ x = N −1 NQ xx N N −1 + N −1 N N −1
• FG IJ • •

H
= N −1 NQ xx N N −1 + Wxx N −1
K
•
= N −1 NQ xx bN + W g N xx
• −1

• • •
= N −1 NQ xx N N −1

or

•
Qδ xδ x = N −1 NQ xx = N + Wxx b g −1
NQ xx (2.66)

© 2003, R.E. Deakin Coordinate Transformations (2003) 2–12

Department of Geospatial Science RMIT

2.4.4. Cofactor Matrix for v

From (2.32), (2.31) and (2.30) we may write the following

v = QA T k
a
= QA T We f − Bδ x f
= QA T We f − QA T We Bδ x
= QA T We f − QA T We B N + Wxx b g −1
t

•
and with (2.29) and the auxiliary N −1 = N + Wxx b g −1

•
v = QA T We f − QA T We B N −1 BT We f (2.67)

v is a function of the observables x and the observations l since f = − F x, l and applying the lawa f
of propagation of cofactors gives

FG ∂v IJ Q FG ∂v IJ + FG ∂v IJ Q FG ∂v IJ
T T

Q vv =
H ∂x K H ∂x K H ∂l K H ∂l K
xx (2.68)

The partial derivatives of (2.67) are

∂v ∂f • ∂f
= Q A T We − Q A T We B N −1 BT We
∂x ∂x ∂x

∂v ∂f • ∂f
= Q A T We − Q A T We B N −1 BT We
∂l ∂l ∂l

∂f ∂f
With = − B and = − A , and with the auxiliary N = BT We B the partial derivatives become
∂x ∂l

∂v •
= Q A T We B N −1 N − Q A T We B (2.69)
∂x

∂v •
= Q A T We B N −1 BT We A − Q A T We A (2.70)
∂l

Substituting (2.69) and (2.70) into (2.68) gives

m r m
Q vv = 1st term + 2 nd term r (2.71)

where

© 2003, R.E. Deakin Coordinate Transformations (2003) 2–13

Department of Geospatial Science RMIT

m1st
r • •
term = QA T We B N −1 NQ xx N N −1 BT We AQ − QA T We B N −1 NQ xx BT We AQ
•

•
− QA T WeBQ xx N N −1 BT We AQ + QA T We BQ xx BT We AQ

m2 nd
r •
term = QA T We B N −1 BT We AQA T We B N −1 BT We AQ
•

•
− QA T We B N −1 BT We AQA T We AQ
•
− QA T We AQA T We B N −1 BT We AQ
+ QA T We AQA T We AQ

The 1st term above is identical to the 1st term of (2.58) which simplifies to (2.59) as

m1 st
r •
term = QA T We B N −1 Wxx N −1 BT We AQ
•
(2.72)

The 2nd term above can be simplified by remembering that AQA T = Q e = We−1 so that after some
manipulation we have

m2 r FG • • • IJ
H
term = QA T We B N −1 N N −1 − N −1 BT We AQ
K
nd

•
− QA T We B N −1 BT We AQ + QA T We AQ

The term in brackets can be expressed as

• • • •
N −1 N N −1 − N −1 = N −1 N − N N − 1 FH IK • •

= N −1
•
cN − bN + W gh N xx
• −1

• •
= − N −1 Wxx N −1

and the 2nd term becomes

m2 nd
r •
term = − QA T We B N −1 Wxx N −1 BT We AQ
•

(2.73)
• −1
− QA T We B N BT We AQ + QA T We AQ

Substituting (2.72) and (2.73) into (2.71) gives the cofactor matrix of the residuals v as

Q vv = − QA T We B N + Wxx b g −1
BT We AQ + QA T We AQ (2.74)

and by inspection of (2.64) and (2.74)

© 2003, R.E. Deakin Coordinate Transformations (2003) 2–14

Department of Geospatial Science RMIT

Q vv = Q − Qll (2.75)

2.4.5. Covariance Matrix Σ xx



Σ xx  = σ 20 Q xx  (2.76)

The estimated variance factor is

v T Wv + δ x T Wxxδ x
σ 20 = (2.77)
r

and the degrees of freedom r are

r = m − u + ux (2.78)

where m is the number of equations used to estimate the u parameters from n observations. ux is the
number of weighted parameters. [Equation (2.78) is given by Krakiwsky (1975, p.17, eqn 2-62)
who notes that it is an approximation only and directs the reader to Bossler (1972) for a complete
and rigorous treatment.]

2.5. Generation of the Standard Least Squares Cases

Depending on the form of the design matrices A and B, and also on whether the parameters are
treated as observables, ie, is Wxx = 0 , there are several different possibilities for the formulation
and solution of least squares problems. The standard cases are listed below.

2.5.1. Combined Case with Weighted Parameters A, B, W, Wxx ≠ 0 b g Av + Bδx = f

The general case of a non-linear implicit model with weighted parameters treated as observables is
known as the Combined Case with Weighted Parameters. It has a solution given by the following
equations (2.30), (2.28), (2.29), (2.26), (2.3), (2.31), (2.32), (2.2), (2.65), (2.52), (2.74), (2.61),
(2.64), (2.77) and (2.78).

b
δ x = N + Wxx g −1
t (2.79)

with N = BT We B (2.80)

t = BT We f (2.81)

© 2003, R.E. Deakin Coordinate Transformations (2003) 2–15

Department of Geospatial Science RMIT

We = Q e−1 = AQA T c h −1
(2.82)

x = x + δ x (2.83)

a
k = We f − B δ x f (2.84)

v = W −1A T k = QA T k (2.85)

l = l + v (2.86)

b
Qδ xδ x = N + Wxx g −1
b
NQ xx N N + Wxx g + bN + W g N bN + W g
−1
xx
−1
xx
−1

(2.87)
= bN + W g
−1
xx NQ xx

b
Q xx  = N + Wxx g −1
(2.88)

Q vv = QA T We AQ − QA T We B N + Wxx b g −1
BT We AQ (2.89)

Qll = Q + QA T We B N + Wxx b g −1
BT We AQ − QA T We AQ (2.90)

Q f f = BQ xx BT + Q e (2.91)

v T Wv + δ x T Wxxδ x
σ =2
0 (2.92)
r

r = m − u + ux (2.93)

Σ δ x δ x = σ 20 Qδ x δ x (2.94)

Σ xx  = σ 20 Q xx  (2.95)

Σ vv = σ 20 Q vv (2.96)

Σ ll = σ 20 Qll (2.97)

Σ f f = σ 20 Q f f (2.98)

© 2003, R.E. Deakin Coordinate Transformations (2003) 2–16

Department of Geospatial Science RMIT

b
2.5.2. Combined Case A, B, W, Wxx = 0 g Av + Bδx = f 0

The Combined Case is a non-linear implicit mathematical model with no weights on the parameters.
The set of equations for the solution is deduced from the Combined Case with Weighted Parameters
by considering that if there are no weights then Wxx = 0 and Q xx = 0 . This implies that x is a
constant vector (denoted by x 0 ) of approximate values of the parameters, and partial derivatives
with respect to x 0 are undefined. Substituting these two null matrices and the constant vector
x = x 0 into equations (2.1) to (2.78) gives the following results.

δ x = N −1t (2.99)

with N = BT We B (2.100)

t = BT We f 0 (2.101)

c h
f 0 = − F x0 , l (2.102)

c
We = Q e−1 = AQA T h −1
(2.103)

x = x 0 + δ x (2.104)

c
k = We f 0 − B δ x h (2.105)

v = W −1A T k = QA T k (2.106)

l = l + v (2.107)

Qδ xδ x = Q xx  = N −1 (2.108)

Q vv = QA T We AQ − QA T We BN −1BT We AQ (2.109)

Qll = Q + QA T We B N −1BT We AQ − QA T We AQ (2.110)

Q f 0 f 0 = Qe (2.111)

v T Wv
σ 20 = (2.112)
r

© 2003, R.E. Deakin Coordinate Transformations (2003) 2–17

Department of Geospatial Science RMIT

r = m−u (2.113)

Σ xx  = Σ δ xδ x = σ 20 Q xx  (2.114)

Σ vv = σ 20 Q vv (2.115)

Σ ll = σ 20 Qll (2.116)

Σ f 0 f 0 = σ 20 Q f 0 f 0 (2.117)

b
2.5.3. Parametric Case A = I, B, W, Wxx = 0 g v + Bδx = f 0

The Parametric Case is a mathematical model with the observations l explicitly expressed by some
non-linear function of the parameters x only. This implies that the design matrix A is equal to the
identity matrix I. Setting A = I in the Combined Case (with no weights) leads to the following
equations.

δ x = N −1t (2.118)

with N = BT WB (2.119)

t = BT Wf 0 (2.120)

c h
f 0 = − F x0 , l (2.121)

x = x 0 + δ x (2.122)

v = f 0 − B δx (2.123)

l = l + v (2.124)

Qδ xδ x = Q xx  = N −1 (2.125)

Q vv = Q − BN −1BT (2.126)

Qll = B N −1BT (2.127)

Qf0f0 = Q (2.128)

© 2003, R.E. Deakin Coordinate Transformations (2003) 2–18

Department of Geospatial Science RMIT

v T Wv
σ 20 = (2.129)
r

r = n−u (2.130)

Σ xx  = Σ δ xδ x = σ 20 Q xx  (2.131)

Σ vv = σ 20 Q vv (2.132)

Σ ll = σ 20 Qll (2.133)

Σ f 0 f 0 = σ 20 Q f 0 f 0 (2.134)

b
2.5.4. Condition Case A, B = 0, W, Wxx = 0 g Av = f

The Condition Case is characterised by a non-linear model consisting of observations only. Setting
B = 0 in the Combined Case (with no weights) leads to the following equations.

k = We f (2.135)

with We = Q e−1 = AQA T c h −1

(2.136)

f = −F l af (2.137)

v = W −1A T k = QA T k (2.138)

l = l + v (2.139)

Q vv = QA T We AQ (2.140)

Qll = Q − QA T We AQ (2.141)

v T Wv
σ 20 = (2.142)
r

r =m (2.143)

© 2003, R.E. Deakin Coordinate Transformations (2003) 2–19

Department of Geospatial Science RMIT

Σ vv = σ 20 Q vv (2.144)

Σ ll = σ 20 Qll (2.145)

© 2003, R.E. Deakin Coordinate Transformations (2003) 2–20