Abstract
Let X (t), -infinity < t < infinity, be a multidimensional strictly stationary continuous time process with randomly missed observations, all of whose moments exist. From the realization X (t), 0 <= t <= N, we get an improved time series Y(t) and investigate its statistical properties. The multitaper spectral density estimator f(XX)((N))(lambda), -infinity<lambda<infinity, of X (t), 0 <= t <= N, is constructed by tapered periodograms. We give the asymptotic statistical properties of f(XX)((N))(lambda). Confidence interval of a spectral density function for X (t) is concluded. We formulate the statistic F-XX((N))(lambda) via f(XX)((N))(lambda) and get its statistical properties.