Pairwise robust estimation of multivariate location and covariance

Pairwise likelihood may be a useful tool for approximating likelihood based inference in the estimation of multivariate location and covariance matrices.The maximum pairwise likelihood estimator is sensitive to the occurrence of outliers, as well as its likelihood counterpart. As the minimum covariance determinant estimator is an effective tool in providing robust inference at general multivariate elliptically symmetric distributions, the employ of pairwise likelihood equations looks challenging. The validity of pairwise based MCD estimation procedures is investigated in the context of mixed linear models and first order autoregression.