Detecting Outliers in the Simple Linear Regression for Children Affected with Leukemia in Mosul City

Abstract

In this study, the problem of outlier detection in “linear regression” analysis is studied using the “median” and “mean” “absolute deviation” about the median. “The mean and standard deviation” are heavily affected by outliers Hence, the outlier detection techniques based on these measures may not correctly identify all outliers in a dataset. However, “the mean absolute deviation about the median”, in combination with the median is sufficiently robust in the presence of outliers and provides a better alternative. The conceptualized method was tested using leukemia patients data and the results indicate that the new method performed better than the methods based on the mean/ standard deviation combination. It is recommended that the median and “mean absolute deviation” about the median be used in detecting outliers in regression analysis due to their inherent potential for increasing the” goodness-of-fit of the “linear regression mode