The text presents a proposal of a parameter estimation method in multivariate time series. The AR(1) model is investigated.
An application to financial data is introduced including a comparison with some outputs of software implementations of standard estimation procedures.