초록 close

We present a diagnostic method for the quasi-likelihood estimators ingeneralized linear models. Since these estimators can be usually obtainedby iteratively reweighted least squares which are well known to be very sen-sitive to unusual data, a diagnostic step is indispensable to analysis of data.We extend the local inuence approach based on the maximum likelihoodfunction to that on the quasi-likelihood function. Under several perturba-tion schemes local inuence diagnostics are derived. An illustrative exampleis given and we compare the results provided by local inuence and deletion.