AS a rule, Ps is determined by means of the result of the prelimi-
nary adjustment or by means of an iterative process. But the ps,
Eq.(14) always takes the value of the weight for the final ad just-
ment.
If Pg con be expressed as a explicit function of the least Squares
estimators in $ , the Eq.(14) will take a more concrete form. For
example, if the weights are determined according to formula:
42,22
= 0/51
: : : 2. 2 ~
(The reason for giving this example is that Psi" 09/8; makes S
reach the minimum mean Square error/ /), then "the concrete
expression for Su is
vs, »———( [81] + „Taf G, Rs; ) i
1 + T T
in which T; is defined as in Eq.(10). In this particular case, the
maximum value V,S; of YS; can be given as follows:
A [1 + 27 A
VoSi 9 mox Ys = pomum O isa =* 415)
It-is the same measure as expressed by formula (7).
The methods with selecting parameters by statistical test can be
considered as a special case of robust methods. Selecting parame-
ters by means of t-test is equivalent to determining the weights by
the following formula:
(° | 71] >to,
6o I| &*.;
Substituting this p
(4) for the case of
si into Eq.(14), we again obtain the expression
Selecting parameters.
From Eq.(11) we see that the biases for s influence the biases for
x in the same formula as in the case of parameter selection. So,
we can obtain the expresSion of the posteriori external reliability
measure for the case of using robust methods simply by substituting
the expression (14) into (8). That is
me + ds; Ps; -1 6
Vp if = pe ———— (Jra| + 68/2) fhjACA'A MM nga, (16)
1 + ds,-Ps;
in which all the symbles have the same meanings as before.
9. CONCLUDING REMARKS
The posteriori reliability measure proposed in the paper reflects
the concrete values of the additional parameters, so is more
accurate in assessing concrete results than the existing reliability
- 372 -