TY - JOUR
T1 - Assessment of Outlier Detection Procedures in Analysis of Regression Model
AU - Adeboye, Azeez AU - James, Ndege AU - Akinwumi, Odeyemi
JO - Pakistan Journal of Social Sciences
VL - 13
IS - 3
SP - 25
EP - 31
PY - 2016
DA - 2001/08/19
SN - 1683-8831
DO - pjssci.2016.25.31
UR - https://makhillpublications.co/view-article.php?doi=pjssci.2016.25.31
KW - Coefficient
KW -diagonal
KW -measures
KW -P-P plot
KW -residual
KW -simulation
KW -unbiased estimator
AB - Five detection of outliers procedures in Multiple regression model are looked into, compared and investigated with a simulated data. The researchers reviewed five outlier detection methods in multiple linear regression model and then compares theirs results by using two criteria of robust diagnostics called the Median Absolute Deviation (MAD) and the Standard Deviation (SD) parameter estimate. Data were generated with 10, 20 and 30% of outliers on X1s, X2s and both X1s and X2s, respectively with different sample sizes (20, 50 and 1 2 1 2 100) to check and compare outliers in the residual space of CovRatio which will flag observations that are influential because of large residual, outliers in the X-space of Hat Diagonal which flags observations that is influential because they are outliers in the X-space, the Dffits shows the influence on fitted values and measures the impact on the regression coefficients. Cooks D measures the overall impact that a single observation has on the regression coefficient estimates and Mahalanobis Distance measures the hat leverage through the means of Mdi.
ER -