A statistic that provides an effective way of detecting an outlier observation. The deletion residual corresponding to the jth of n observations is calculated by comparing its value with the corresponding fitted value based on the model with the parameters estimated from the remaining (n−1) observations. See regression diagnostics.