Kybernetika 28 no. 3, 213-226, 1992

Nonnegative multivariate \operatorname{AR}(1) processes

Jiří Anděl


Conditions for nonnegativity of a p-dimensional AR(1) process $X_t = U X_{t-1} + \e_t$ are investigated in the paper. If all the elements of the matrix U are nonnegative, a new method for estimating U is proposed. It is proved that the estimators are strongly consistent. Small-sample properties of the estimators are illustrated in a simulation study.


62F12, 62M09, 62M10