Abstract: This paper deals with the deduction of two methods respectively used for the elimination of the "gross-error point" (i.e.all the observations with gross errors connected to the point) and the observations with gross errors. The deduction is drawn from the concept of the pseudo-weight and the formulas of filterling and collocation. In addition, the author makes a simplified deduction of the formula for the sequencial least squares collocation with limited memory and develops its formula for computational accuracy.A simple levelling net-work is referred to as an example of illustrating the application of the mentioned methods. In the closing part, the author makes some suggestions concerning the further studies of pseudo-weight.