The article deals with the behavior of reduced scalar estimates in the presence of systematic errors in the observational data. The proposed procedure with a different method of forming the reduction coefficient. A quasi-optimal variant of the compression parameter formation is considered. Simulation results for different conditions of application of the proposed algorithms are presented. Currently, one of the ways to improve the accuracy of the formation of the time scale in solving the problems of frequency-time customer support is the averaging of the readings of several generators. At the same time, this approach, as shown in the theory of statistical estimation, is effective for parrying the random component of the error of the estimated process. However, for frequency generators random error can be effectively compensated for a long range of observations, but the systematic component - frequency drift - is a serious problem, which can be eliminated by averaging only under certain conditions. Therefore, the article proposes a version of the reduced estimate, effective, as shown, to parry the departure of the time scale by introducing a shift in the implementation of compression, defined by the reduction procedure. The conditions in which the degree of the achieved positive effect has a practical sense are considered.
Statistical estimation of word acquisition with application to readability prediction
