Dear Seppo;
My coauthors and I have been using "Unconfounded Mechanism" for survey
sampling situations, which corresponds to MAR under a super-population
set-up. This term was also coined by Rubin (1983) but we used it in a little
bit narrower sense. Our framework is conditional on a given sample, that is,
the unconfounded response mechanism is that the response probability is
independent of auxiliary variables observed for the whole sample (for more
detail see Lee, Rancourt, and Sarndal, 1994, JOS).
Best,
Hyunshik Lee, Ph.D.
Westat
1650 Research Blvd.
Rockville, MD 20850
E-mail: [EMAIL PROTECTED]
Tel: 301-610-5112
Fax: 301-294-2034
-----Original Message-----
From: Laaksonen Seppo [mailto:[EMAIL PROTECTED]]
Sent: Thursday, March 29, 2001 5:39 AM
To: [EMAIL PROTECTED]
Subject: IMPUTE: MAR
I do not like about the term MAR, missing at random. Of course, when it has
been defined, there are no problems. But the direct interpretation of that
term
is confusing, since missingness is not random in this case but conditionally
in
some sense. The term should be something like missing at random
conditionally
(MARC) or MAR according to covariates. I am not fully satisfied to those
terms.
What do you prefer?
Best regards
Seppo Laaksonen
