Yuri Petrovich Dranitsa
The Murmansk State Technical University,
Russia, Murmansk
Andrei Yurevich Dranitsa
Closed Joint-Stock Company "Lanit",
Russia, Moscow
Olga Vasilevna Alekseevskaya
Closed Joint-Stock Company "Lanit",
Russia, Moscow
The fast algorithm of construction of the linear autoregression
for vector time series which are non-stationary relative to 2nd order statistics has been developed.
The method is based on Toeplitz's recursive procedure and does not require the knowledge
of the correlation matrix of data. The comparative analysis of the technique with the classical
decision of this problem has been performed.
The method can be used for the estimation of dynamic characteristics of vector time series in non-stationary case.
Keywords:
Autoregression equation, vector time series, fast algorithms, correlation function, the method of maximal entropy,
function of partial correlation.