The minimum weighted covariance determinant estimator revisited