“DFBETAS” is short for “difference in betas”; “beta” is another way to say coefficient. It is a deletion diagnostic where the influence of each run on a coefficient estimate is measured by removing each run and recomputing the coefficients.
DFBETAS are computed separately for each term in the model. A DFBETA statistic is the number of standard deviations that the jth coefficient changes when the ith run is removed from the analysis.
A run is considered influential on a coefficient when its DFBETAS value is outside of the limits.
The limits are computed as…
Where n is the number of runs in the design.
A high DFBETAS indicates that the coefficient for that term is changing because of influence from the run.
Never ignore a run just because the diagnostics plots indicate it may be a problem. Verify that the data is wrong in some way before ignoring it.