Consider a bivariate data set and imagine that the data are projected onto the x-axis and then checks for outliers are made using the MADmedian rule. The same is done when projecting the data onto the y-axis and both times no outliers are found. Why might the conclusion of no outliers be unsatisfactory?