DFITS tells how unusual an observation in a data set is by combining the leverage and the studentized residual.  More specifically, it is the difference between the fitted (predicted) values calculated with and without the i^th observation.  Having unusual data can have an influence upon the regression result, so it is necessary to recognize any data that might skew the results.  An observation is considered to be unusual, for small to medium data sets, if the absolute value of the DFITS value calculated is greater than 1.  We shall demonstrate this with example.

The DFits can be calculated by hand, using Minitab and using Excel.  Calculating the formula by hand is difficult but possible, while using Minitab and Excel are relatively easy.

Regression Tutorials Menu  STATS @ MTSU   Dictionary