The results include several types of residuals to assess the adequacy of the Cox proportional hazards model. For formulas to calculate the residual for each observation, go to Methods and formulas for the regression diagnostics in Fit Cox Model with Fixed Predictors only.
For data in counting process format with a variable that identifies the rows for each case, Minitab Statistical Software calculates per-subject residuals. The per-subject residual is the sum of the residuals from all the rows for that subject. If the analysis does not include a variable that identifies the rows for each case, then Minitab calculates a separate residual for each row. Separate residuals for each row are not necessarily per-subject residuals.
The calculation for the per-subject deviance residuals differs from the calculation given for the fixed predictors case.
Deviance residuals
is the Martingale residual for subject
i, which is the collapsed residual for the subject row residuals in
this case. Additionally, 
is the total event counts for the subject, which is equivalent to the sum of
the events for all of the rows for that subject. The deviance residual is
missing in the following conditions:
