MULTIPLE IMPUTATION IN MIXTURE-MODELS FOR NONIGNORABLE NONRESPONSE WITH FOLLOW-UPS

One approach to inference for means or linear regression parameters wh en the outcome is subject to nonignorable nonresponse is mixture model ing. Mixture models assume separate parameters for respondents and non respondents; implementation by multiple imputation consists of repeate dly filling in missing values for nonrespondents, estimating parameter s using the filled-in data, and then adjusting for variability between imputations. We evaluated the performance of this scheme using simula ted data with a 25% sample of nonrespondents followed up. We conclude that it provides a generally satisfactory and robust approach to infer ence for means and regression parameters in this case, although a grea ter number of imputations may be required for good performance compare d to the number required for estimation when nonresponse is ignorable.