nt 
AV 
MONTSALV ENS 
' ; J 
1 
\ 
| 
. \ | * 
ES OL (b à [srr F 
- 
Scale 
| 
[/ 
Y 
y | | —9. Map 
/ © ,$ Qys , Mr |. gllipse/ Arrow 
Fig. 3 Shift between epoch 1 and epoch 2 with a gross error 
non-distinguishable model errors on the adjusted results ise 
In order to be consistent with the reliability theory of 
Baarda a separability multiplying factor is introduced. 
2. THEORY OF SEPARABILITY 
2.1 Statistical Test under two alternative hypotheses 
Let be 
^ "y 2 1 
E(1) » Ax ( “= true value) D(1) = oo P (1) 
the linear model ,whereby 
1=n X 4 vector of observations, 
A =n x u design matrix with rk(A)zu, 
x = u x 1 vector of unknowns, 
P =n x n weight matrix of observations. 
The null hypothesis 
. 5 ^u 
Hio E(I/H,) s AX (2a) 
is to be tested against two alternative hypotheses 
Hai: E(1/Haj) = E(1/Hg) * H,v$4 (i= 1.2.) , (2b) 
where Hi =n x pj matrix with rk( Hy )= Pi, 
pix 1 vector of additional parámeters. 
^, 
VS; 
- 189 -