Data evaluation efficiency on the base of a nonuniform sample

The comparative analysis of the efficiency of different methods used for determining the estimates of parameters obtained as a result of experimental data processing in the presence of incomplete or rough a priori information is made. The asymptotic dispersion of various estimates obtained by the methods of sample mean, sample median, sample truncation as well as by the Huber method and the method of maximum-likelihood fit are compared for a nonuniform sample. The effect of bias spread on the asymptotic dispersions is also estimated for the case when data processing is made without taking into account this bias. The conclusion is drawn that the best results are obtained by the method of sample truncation and by the Huber method