UNBIASED FREE NET ADJUSTMENT

Erik W. Grafarend, Burkhard Schaffrin

Survey Review · 1974 · 72 citations · 1 references

Concepts

Abstract

Using generalized inverse matrices it is proved that in general solutions for free net adjustments characterized by singular normal equation matrices are biased, therefore nonestimable. For any type of defect (origin, orientation, scale) in a free geodetic net operators are generated which transform biased parameters into unbiased ones. The number of linearly independent estimable parameters is determined by the rank of the coefficient matrix within observation equations. Based on the Moore–Penrose or pseudoinverse, which has the least-squares minimum norm property, examples are computed which show how contradictory the results with bias are: Absolute heights are predicted without any fixed height, the scale of a purely angular net is determined. Therefore, it is plausible to adjust for parameters which are a priori known to be unbiased estimable. For both ways (free net adjustment with singular normal equation matrix plus transformation and direct approach for estimable parameters) conditions for the "commutativity" are given.

References

1