Home / DOE / Additional Information / Binomial Logistic Residuals

Understanding Binomial Logistic Residuals

Residual analysis is a method of regression diagnostics. Residuals can help you identify influential points (potential outliers), lack of fit issues, and other patterns. The ultimate goal is to improve the model.

Leverage

Leverages (h~i~) for logistic regression are similar to those in Ordinary Least Squares; however, they are calculated differently due to the nature of logistic regression. Observations with higher leverage values exert more influence than other observations.

Delta Chi-Squared

Delta Chi-Squared (Deletion Chi-Squared, Delta Pearson Chi-Squared, Deletion Pearson Chi-Squared) measures the change in the Pearson chi-squared statistic if the ith covariate pattern is removed from the model. Observations that do not fit the model well have large Delta Chi-squared values.

And the Pearson residual (r~i~) is

Delta Deviance

Delta Deviance is a measure of the change in the deviance statistic when the all of the observations with the ith covariate pattern are removed.

Delta Beta

Delta Beta (Delta displacement) is a measure of the change in the coefficients when all of the observations with the ith covariate pattern are removed. Larger values indicate points with more influence on the model.

Standardized Delta Beta

Standardized Delta Beta (Standardized Deletion Displacement) measures the change by deleting all observations with the ith covariate pattern. Larger values have strong influence on regression coefficients.