\(R\)-estimation of the parameters of autoregressive [AR(\(p\))] models (Q2366756)

The authors consider \(R\)-estimation of a subset of the parameters of a stationary linear \(\text{AR}(p)\) when the complementary subset may be redundant. They compare the unrestricted and restricted \(R\)-estimators with preliminary-test and shrinkage \(R\)-estimators. Using asymptotic distributional risk as the criterion, the latter two estimators are found to be robust with respect to the complementary subset. The restricted estimator may dominate when this subset actually is redundant but is behaves poorly otherwise. More detailed results are derived.