I forget what EM is (or maybe I never knew). Nonetheless, my humble opinion is that one should not impute in such a case. Of course there is the possibility of bias if you dump the people who refuse to answer. How any kind of imputation would eliminate any such bias is beyond me. Besides, if you impute your analysis proceeds as if you have a full data set. Unless you have used some imputation method that adds to the variance, you will have a systematic bias (underestimate) of your standard errors. There are some considerations: Are your data weighted? What percent are non-randomly missing these items? What method of analysis is planned? Is this for your own use or for a public use file? What are your options in SPSS for estimating standard errors?
John Hall Senior Sampling Statistician Mathematica Policy Research 600 Alexander Park Princeton, NJ 08540 phone (609) 275-2357 fax (609) 799-0005 email [email protected] -----Original Message----- From: Donald Baken [mailto:[email protected]] Sent: Tuesday, June 19, 2001 6:21 PM To: [email protected] Subject: IMPUTE: non-random missing data I am a new subscriber to this group so I hope that this question is not too simple. I have looked at the archives but it does not seem to be covered. I am using SPSS for my analysis. I have both MCAR and non-random missing data. The non-random missing data comes from questions about religion where some people have just refused to answer any of the questions. I am planning to use EM to impute the missing data for the items with MCAR data. The non-random missing data is a bit trickier. I don't want to dump the items with non-random missing data because they are important. I can't dump the people who didn't answer the items because that would introduce bias. Although the SPSS manual suggests that EM is not suitable for data which is not missing at random I see that as my best option. What do people normally do in this sought of situation? Thanks Don Baken Don Baken Graduate Assisstant School of Psychology Massey University New Zealand
