Internal Reliability Test. Baarda
Internal reliability is taken to mean the detection and control of any possible coarse errors in the observables. In those way it is possible to ascertain network sensitivity to any coarse errors.
To study the internal reliability of the network we will use the following parameters:
Redundancy of each observable
Baarda Parameter
Minimum detectable error
Redundancy of each observable
The redundancy of an observable is a dimensionless parameter and they show us the good or bad controlled by said observable. The expression which allows us to calculate the number of redundancies of an observable is:
where:
ri = Redundancy of any observable.
pi = Weight of an observable.
qi = Residue cofactor a posteriori of the observable.
Baarda Parameter
This parameter depends on the level of significance end the test power set for the network. A significance level of 99.99 % - = 0.001 - has been set and a test power for coarse error detection of 80% - = 0.2 – as they have been assumed as more suitable values for the efficiency of this test. The Baarda parameter is obtained from the following expression:
where:
Vi = Residues vector
σRi= Typical deviation of the residues
The Baarda parameter, along with the minimum detectable error, is some of the coefficients used to reject or eliminate an observable. What's more, this parameter allows the gross errors introduced into the network to be controlled.
In this way an observable will be rejected when the value of the Baarda parameter is greater than the percentage point (3.2905) set for the significance level.
Minimum detectable error
The minimum detectable error for an observable is obtained from the following expression:
where:
δ = Translation parameter
As we can observe, this parameter is determined in line with the translation parameter which is corresponded to the displacement produced in the Gauss campaign owing to the coarse error.
Essentially, these parameters are determined for a significance level and for a data test power, ensuring that a correct observable will not be rejected, with a probability of 99.9 % and any such possible gross errors will be detected with a test power of 80 %, meaning that 20% thereof could be introduced into the adjustment.
TcpMDT
