Many thanks to Venita and Jason. I was sure I remebered that but didn't want to spout off without confirmation.
Paul R. Swank, Ph.D. Professor, Developmental Pediatrics Director of Research, Center for Improving the Readiness of Children for Learning and Education (C.I.R.C.L.E.) Medical School UT Health Science Center at Houston -----Original Message----- From: [email protected] [mailto:[email protected]] On Behalf Of [email protected] Sent: Wednesday, June 28, 2006 9:57 AM To: [email protected] Subject: RE: [Impute] Complete case analysis Hi Paul, Schafer (1997), Little and Rubin (2002), and others have all noted that casewise deletion is based on MCAR. When complete cases are removed, we have no way to control the missingness mechanism. However, if the discussion was on multiple regression, then the discussion is a bit expanded: In multiple regression, if the missingness on the independent variables doesn't depend on the missingness on the dependent variable (and all other standard assumptions are met), then data which are just MAR are almost always appropriate for casewise deletion (Allison, 2002). Moreover, logistic regression is even more relaxed for the use of casewise deletion. The missingness mechanism may depend on either a dependent variable or the independent variables, just not both, in order to obtain unbiased slopes and standard errors; the intercepts will, however, be biased (Vach, 1994). The three primary problems with casewise deletion, beyond needed to be MCAR (and somewhat related to this assumption of MCAR) are: reduced power, skewed standard errors, and lack of generalizability when even a moderate amount of missingness is present. Hope this helps, Jason ____________________________________ Jason C. Cole, PhD Senior Research Scientist & President Consulting Measurement Group, Inc. Tel: 866 STATS 99 (ex. 5) Fax: 818 905 7768 7071 Warner Ave. #F-400 Huntington Beach, CA 92647 E-mail: [email protected] web: http://www.webcmg.com ____________________________________ -----Original Message----- From: [email protected] [mailto:[email protected]] On Behalf Of Swank, Paul R Sent: Wednesday, June 28, 2006 7:42 AM To: [email protected] Subject: [Impute] Complete case analysis Someone on another list made the argument that while imputation and Mlsolutions to missing data problems require the assumption of missing at random, complete case analysis does not. I was under the assumption that complete case analysis, or listwise deletion of missing data required MCAR (Missing completely at random) assumption. Any comments? Paul R. Swank, Ph.D. Professor, Developmental Pediatrics Director of Research, Center for Improving the Readiness of Children for Learning and Education (C.I.R.C.L.E.) Medical School UT Health Science Center at Houston _______________________________________________ Impute mailing list [email protected] http://lists.utsouthwestern.edu/mailman/listinfo/impute _______________________________________________ Impute mailing list [email protected] http://lists.utsouthwestern.edu/mailman/listinfo/impute
