Premium
ESTIMATION OF SPATIAL ARMA MODELS
Author(s) -
Huang D.,
Anh V.V.
Publication year - 1992
Publication title -
australian journal of statistics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.434
H-Index - 41
eISSN - 1467-842X
pISSN - 0004-9581
DOI - 10.1111/j.1467-842x.1992.tb01066.x
Subject(s) - identifiability , estimator , autoregressive–moving average model , mathematics , estimation , consistency (knowledge bases) , series (stratigraphy) , estimation theory , strong consistency , time series , algorithm , computer science , econometrics , statistics , autoregressive model , paleontology , geometry , management , biology , economics
Summary Spatial ARMA models are considered using the nonsymmetric half plane ordering on a lattice of data. A method is given for the estimation of the orders and the coefficients of such models under an identifiability condition and the condition that the beat linear predictor is the best predictor in the mean square sense. Under these conditions, the strong consistency of the estimators ia established. The usual methods for ARMA modelling in Time Series Analysis require estimation of the innovations. The method of this paper introduces an inveree model complementary to the original model so that the estimation of the innovations is avoided. This leads to a substantial reduction in the computational complexity in the two‐dimensional case.