For each observation, the deviance residual measures the change in deviance from the full to the null model. Each deviance residual is the square root of the deviance goodness-of-fit statistic for each cell (or distinct radius value). The deviance residual is

Where the sign is positive if the observed yi is higher than the estimated mean and negative if the observed yi is less than the estimated mean. Repeat Question 27 but use the deviance residual instead of the Pearson residual.

(7.24)