Robust estimation of the shift parameter under a bounded noise variance (Q761407)

The problem of robust estimation of the shift parameter (the center of the distribution) for symmetric densities is considered. Special attention is given to problems with bounds on the variance. The obtained solutions are used in the case of densities of the mixture type. One shows that the character of the solution is determined by the magnitude of the bound: For small variances the Gaussian-type densities and their corresponding data processing algorithms by the least-squares method are optimal; for large variances the Laplace density and the algorithm of the least-moduli method are optimal.