On the efficiency of spectral density estimates for a stationary process. II

The paper is the continuation of [1]. We present an asymptotical analysis and comparison of different statistics for spectral densities of stationary processes with finite sample length. The comparison of the mean square errors in done with respect to the local properties of a spectral density as well as to the strong jitters and noise in the neighbouring frequences. The paper provides also a new type of statistics which is nearly insensible to all sorts of interference.