Which statistic is the difference between the adjusted predicted value and the original predicted value for a case?

Which statistic is the difference between the adjusted predicted value and the original predicted value for a case?

• A. DFBeta
• B. Discrete variable
• C. DFFit
• D. Direct effect

Answer: (C)

In regression diagnostics, a key idea is how much a single observation can sway the model’s prediction for that same observation. This statistic does exactly that: it compares the predicted value for a case when all data are used to the predicted value for that case when that case is left out and the model is refitted. The difference between these two predictions is the DFFIT. A large absolute DFFIT means the case has a strong influence on its own predicted value, flagging a potential outlier or high-leverage point. This is different from statistics that measure impact on the model’s coefficients (DFBETAs) or a standardized version of the same idea (DFFITS). The other term listed doesn’t describe this leave-one-out change in the predicted value.

In regression diagnostics, a key idea is how much a single observation can sway the model’s prediction for that same observation. This statistic does exactly that: it compares the predicted value for a case when all data are used to the predicted value for that case when that case is left out and the model is refitted. The difference between these two predictions is the DFFIT. A large absolute DFFIT means the case has a strong influence on its own predicted value, flagging a potential outlier or high-leverage point. This is different from statistics that measure impact on the model’s coefficients (DFBETAs) or a standardized version of the same idea (DFFITS). The other term listed doesn’t describe this leave-one-out change in the predicted value.