10.3 Precision and reliability

BINGO derives the standard deviations for all adjusted unknowns as a criterion of precision. These standard deviations are valid if the selected mathematical model is correct and if a normal distribution of observations can be assumed. These prerequisites are fulfilled when there are neither gross nor systematic errors, and if the weight functions have been correctly selected. This is guaranteed for a sufficient number of redundant observations if the tests explained in Section 10.2 have been performed.

Another important criterion for evaluation of a network is the inner reliability. For this purpose, BINGO computes the redundancy component for all observations. A redundancy component of 0.6, for example, states that 60% of a possible observation error appears in the corresponding residual v. The remaining 40%, however, falsifies the unknowns of the block. If the redundancy component is large, an error in the observation can be easily detected. If the redundancy component is small, however, errors are difficult to isolate.

BINGO will print a warning message if redundancies are < 0.001. Such observation is uncontrolled. It does not make sense to compute w and ∇. (See Section 2.3.2.3).

The net configuration for a planning a network with auxiliary simulation computation should be selected with redundancy components not less than 0.25. Observations with redundancy components greater than 0.75 may be easily dispensed removed, because the influence to the result is very low.