You are right, Susanne.
Bayesianly proper imputation does not necessarily imply proper imputation.
For example, the multiple imputation method using regression model proposed
by Schenker and Welch (1988) is Bayesianly proper, but not proper in the
sense of Rubin (1987). But, once it is Bayesianly proper, some of the
conditions of proper is automatically satisfied.
Jae
---------------------------------------
Jae-Kwang Kim, PhD
Senior Statistician
Westat
1650 Research Blvd.
Rockville, MD 20850
(301)610-5586
