Detection of Influential Observations in Spatial Regression Model Based on Outliers and Bad Leverage Classification

Influential observations (IOs), which are outliers in the x direction, y direction or both, remain a problem classical regression model fitting. Spatial models have peculiar kind of because they local nature. also not free from effect influential observations. Researchers adapted some techniques to spatial and obtained satisfactory results. However, masking or/and swamping remains stumbling block for such methods. In this article, we obtain measure Studentized prediction residuals that incorporate information on dependent variable residuals. We propose robust diagnostic plot classify into regular observations, vertical outliers, good bad leverage points using classification based potentials, refer as ISRs?Posi ESRs?Posi. Observations fall categories referred IOs. Representations measures general presented. The commonly used diagnostics, Cook’s distance, is compared methods, Hi2 (using non-robust measures), our proposed ESRs?Posi plots. Results simulation study applications real data showed Hsi12 Hsi22 were very successful detecting suffered effect, models. Interestingly, results plot, followed by was classifying correct groups, hence correctly