The statistic that represents the difference between a case's predicted value from the model when that case is included and when it is excluded is called what?

The statistic that represents the difference between a case's predicted value from the model when that case is included and when it is excluded is called what?

• A. Cook's distance
• B. Leverage
• C. Standardized residual
• D. DFFit

Answer: (D)

This is about how much a single observation changes its own predicted value when it’s removed. DFFIT measures the difference between a case’s predicted value with all data and the predicted value for that case when that case is deleted. In other words, it directly captures the influence of that observation on its own fitted value. This differs from other diagnostics: leverage looks at how far the case’s predictor values are from the center of the predictor space, not the actual effect on predictions; standardized residuals indicate how unusual the observed response is given the model, not how much the prediction would change if the case were omitted; Cook’s distance assesses the overall influence of a case on all fitted values collectively rather than on the case’s individual prediction.

This is about how much a single observation changes its own predicted value when it’s removed. DFFIT measures the difference between a case’s predicted value with all data and the predicted value for that case when that case is deleted. In other words, it directly captures the influence of that observation on its own fitted value. This differs from other diagnostics: leverage looks at how far the case’s predictor values are from the center of the predictor space, not the actual effect on predictions; standardized residuals indicate how unusual the observed response is given the model, not how much the prediction would change if the case were omitted; Cook’s distance assesses the overall influence of a case on all fitted values collectively rather than on the case’s individual prediction.