I wish I knew the answer to this, but it is something I am really interested in. Almost all current imputations use missing at random assumptions so don't need to make this decision.
I have a few applications that I have tried this out for. I gave a paper about it at a conference this year that I could forward my slides to you if you are interested. Do you have a data set that you want apply these methods to? Prof Gillian Raab Applied Statistics Group School of Mathematics and Statistics Napier University, Edinburgh tel 0131 455 2674 fax 0131 455 2651 -----Original Message----- From: [email protected] [mailto:[email protected]] Sent: 14 December 2000 10:41 To: [email protected] Subject: IMPUTE: What is the best appoach to deal with categorical data subject to nonignorable nonresponse: selection models or mixture models? If there is no approach, how can be determined which model is the best approach to the data?
