Experiences about Detecting Multiple Outliers in Heteroscedastic Regression Models