The computation is based on the linear or linearised model of
by least squares, possibly with constraints (restrictions) for parameters.
l+v=Ax, BTx=b, vTPv = min!
- l is the known n-vector of (linearized) observations
- v is the unknown n-vector of residuals
- A is the known n×u-design matrix of the adjustment model
- x is the unknown u-vector of (linearized) parameters
- B is the known u×m-matrix of the constraints
- b is the known m-vector of (linearized) constraints
- P is the known n×n-weight matrix of the observations
The model contains as a special case also the model of a statistical sample, e.g. in the form of
. For this case let
u=1,m=0.