From itpeters at indiana.edu  Wed May 28 18:49:10 2014
From: itpeters at indiana.edu (Isaac Petersen)
Date: Wed, 28 May 2014 12:49:10 -0400
Subject: [Mediation-information] Handling missing data
Message-ID: <CAPBn5Xt9sj=FRKxLA0SGuyRykah2XB5cnfEGZ0FtQmn45pjhxA@mail.gmail.com>

Are there plans for the mediation package to handle missing data using all
of the available information (e.g., full information maximum likelihood)
like the bmem package?  Maximum likelihood handling of missing data would
be preferable in my situation to listwise deletion or multiple imputation.
 Bmem doesn't handle clustering, however, so I'm stuck between the two
packages.  If there aren't current plans to implement maximum likelihood
handling of missing data, I hope you'll consider it.  It would greatly
expand the utility of the software so that one doesn't have to rely on
listwise deletion (which is generally ill-advised) or on external packages
for multiple imputation (which works best with larger sample sizes with low
rates of missingness, and can be a headache when combining results across
many data sets).

Thanks!
-Isaac
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.r-forge.r-project.org/pipermail/mediation-information/attachments/20140528/eb1fe4cd/attachment.html>

From dtingley at gov.harvard.edu  Wed May 28 20:53:02 2014
From: dtingley at gov.harvard.edu (dustin tingley)
Date: Wed, 28 May 2014 14:53:02 -0400
Subject: [Mediation-information] Handling missing data
In-Reply-To: <CAPBn5Xt9sj=FRKxLA0SGuyRykah2XB5cnfEGZ0FtQmn45pjhxA@mail.gmail.com>
References: <CAPBn5Xt9sj=FRKxLA0SGuyRykah2XB5cnfEGZ0FtQmn45pjhxA@mail.gmail.com>
Message-ID: <CAPcL1f0NUh6iJ+kYmxd_h8RpBV=otiO5XcgwC9hdUEctEaz1gg@mail.gmail.com>

Thanks Isaac,
We do not have plans for this in short term, but it is good to know about
the bmem package. That may help as we role this feature out in the future.
I agree this would be nice to have.
thanks,
Dustin


Dustin Tingley
Paul Sack Associate Professor of Political Economy
Government Department
Harvard University
http://scholar.harvard.edu/dtingley

Director, Undergraduate Research Scholars, Institute of Quantitative Social
Science
Director, Program on Experience Based Learning in the Social Sciences
http://projects.iq.harvard.edu/peblss/
Contribute to ABLConnect
http://ablconnect.harvard.edu/
Contribute to the Research Reservoir:
http://projects.iq.harvard.edu/undergradscholars/pages/research-resources


On Wed, May 28, 2014 at 12:49 PM, Isaac Petersen <itpeters at indiana.edu>wrote:

> Are there plans for the mediation package to handle missing data using all
> of the available information (e.g., full information maximum likelihood)
> like the bmem package?  Maximum likelihood handling of missing data would
> be preferable in my situation to listwise deletion or multiple imputation.
>  Bmem doesn't handle clustering, however, so I'm stuck between the two
> packages.  If there aren't current plans to implement maximum likelihood
> handling of missing data, I hope you'll consider it.  It would greatly
> expand the utility of the software so that one doesn't have to rely on
> listwise deletion (which is generally ill-advised) or on external packages
> for multiple imputation (which works best with larger sample sizes with low
> rates of missingness, and can be a headache when combining results across
> many data sets).
>
> Thanks!
> -Isaac
>
> _______________________________________________
> Mediation-information mailing list
> Mediation-information at lists.r-forge.r-project.org
>
> https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/mediation-information
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.r-forge.r-project.org/pipermail/mediation-information/attachments/20140528/5c45eac8/attachment.html>