Multivariate one-sample location problem for clustered data is considered from a nonparametric viewpoint. We develop the spatial median and its affine equivariant version as companion estimates to the affine invariant sign test proposed recently by Larocque (2003). The asymptotics of the proposed estimates are extended to dependent data, and their limiting as well as finite sample efficiencies for multivariate t-distributions are explored. We learn that the efficiency of the spatial median suffers less from intracluster dependencies than the mean vector.