The optimal value of the window length in singular spectrum analysis (SSA) is considered with respect to the concept of separability between signal and noise component, from the theoretical and practical perspective. The theoretical results confirm that for a wide class of time series of length N, the suitable value of this parameter is median {1, . . . ,N}. The results of both simulated and real data verify the effectiveness of the theoretical results. The theoretical results obtained here coincide with those obtained previously from the empirical point of view.
