Least Absolute Regression Residuals: Analyses of Split-Plot Designs