I like the missing at random conditionally term - MARC

For informative missing or non-ignorable non-response 
let's also come up with a better
term.  I seldom see anything that is ignorable in statistics.
One "straw man" would be missing at random conditionally on
unknowns or MARCU; a better one might be missing dependent
on unknown response or MDUR.  I'm sure others can improve
on this.  It's fun coming up with new names; getting them
used is another matter.

-Frank Harrell


Werner Wothke wrote:
> 
> Rod:
> 
> Sorry, I have to side with Seppo Laaksonen on this one. I teach missing
> data handling to applied social scientists as part of my structural
> equation modeling workshops. In my experience, the MAR-MCAR distinction
> is confusing at best. For non-statisticians, these two terms simply
> *sound* too similar to make a meaningful distinction. In addition,
> people's common-sense understanding of the term "missing at random"
> is really MCAR. This is reflected in Seppo's desire to change the
> "MAR" terminology to "missing at random conditionally." In other words,
> the MAR-MCAR terminology is somewhat of an obstacle to teaching
> ML-based missing data methods to non-statisticians.
> 
> I do not mean to diminuish Don Rubin's eminent contributions to the
> missing data field in any way--quite the contrary. On the other hand,
> after 27 years of MAR-MCAR it is really time to consider replacing that
> terminology with one more easily understood by applied researchers.
> Yours and Don's assistance in this matter would help a lot.
> 
> Cheers,
> 
> Werner Wothke, Ph.D.
> 
> Rod Little wrote:
> >
> > Dear Seppo: you are not the first to make such a comment. I'd say that MAR
> > is one of the few terms in the area I like, since it at least has a
> > well-defined meaning! It was originally coined in Rubin's 1976 Biometrika
> > paper, and the rationale was that people were doing analyses ignoring the
> > missing data mechanism that had an implied assumption, and he defined MAR
> > to be that implied assumption. For likelihood inference that assumption is
> > MAR. Since there is another reasonable term for the "randomness" you are
> > talking about, namely missing completely at random, I see no real problem.
> > There is an strong argument that people who write the seminal article
> > should get to define terms, and I think trying to change a definition is
> > asking for confusion.
> >
> > Concerning other terms, is there an established definition of
> > "informatively missing"? I'd be interested in comments on this. Best, Rod
> >
> >  On 29 Mar 2001, Laaksonen Seppo wrote:
> >
> > > 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
> 
> --
> ***********************************************************************
> *                                                                     *
> *  SmallWaters Corporation         phone: USA-773-667-8635            *
> *  1507 E. 53rd Street, #452         fax: USA-773-955-6252            *
> *  Chicago, IL 60615              e-mail: [email protected]        *
> *  USA                          web page: http://www.smallwaters.com  *
> *                                                                     *
> ***********************************************************************

-- 
Frank E Harrell Jr              Prof. of Biostatistics & Statistics
Div. of Biostatistics & Epidem. Dept. of Health Evaluation Sciences
U. Virginia School of Medicine  http://hesweb1.med.virginia.edu/biostat

Reply via email to