| Abstract: | Regression diagnostics detect outliers. The purpose of this study is to analyze the effect of measurement error on the detection process of the regression diagnostic, the hat diagonals. The hat diagonals are analyzed according to four cases: the error-free and contamination-free case (Model 1), the measurement error case (Model 2), the outlier case (Model 3), and the combined measurement error and outlier case (Model 4). The analysis leads to the limiting probability distribution of (n-1)hᵢᵢ under Model 4. This distribution is simulated for sample 11 sizes, N, equal to 20, 40, and 60. |
|---|