Regarding doing MI outside WinBugs, see
A Note on Bayesian Inference After Multiple Imputation Xiang Zhou, Jerome P. Reiter. The American Statistician. May 1, 2010,
64(2): 159-163.
There is a "good" way of doing this...
Also, this issue came up before. Look for past post with subject heading: " MI and Bayesian Analysis: Summary"
Cheers,
Samuel H. Field
Statistician Investigator
Frank Porter Graham Child Development Institute
University of North Carolina - Chapel Hill
-----Original Message-----
From: (The BUGS software mailing list) [mailto:[log in to unmask]] On Behalf Of Jonathan Christensen
Sent: Thursday, November 04, 2010 11:30 AM
To: [log in to unmask]
Subject: Re: [BUGS] imputing missing values - summary of replies
Amalia,
You're better off not imputing the missing data separately. When BUGS
imputes the missing data during the analysis, it takes the uncertainty
associated with the missing data into account. Finding point estimates
for the missing data and then running the analysis from that will give
misleading results.
Jonahtan
On Wed, Nov 3, 2010 at 6:43 PM, M Jerison <[log in to unmask]> wrote:
> Hi,
> Thank you to everyone who replied to my question (I am wondering whether
> anyone can tell me the procedure Bugs uses to impute missing data. Also, is
> there a way to first impute the missing data and then run a Bugs program on
> it?).
>
> Here are some of the answers I received.
>
> This document has a helpful introduction to missing values in Winbugs:
>
> http://psc.maths.lancs.ac.uk/shortCourses/notes/missingdata/MissingBayesian.beamer.pdf
>
> The following authors have done a lot of research in the area of missing
> values analysis:
> G. King, P. Allison, J. Graham, D. Rubin, J. Schafer
>
> (fom Sotiris Adamakis)
>
> This powerpoint is also useful for learning about missing values:
>
> home.uchicago.edu/~grynav/bayes/ABSLec18.ppt
>
> (from Kenneth Elgersma)
>
>
> George Woodworth wrote:
>
> Bugs "imputes" different values for each missing observation in each
> MCMCupdate, which is exactly what it does for parameters. The closest
> frequentist analog is random imputation. Values are "imputed" from the
> conditional distribution of the missing observation given the observed
> data and the parameter values in the current update.
>
> You can set the Sample Monitor Tool to monitor the nodes corresponding
> to missing observations. For example if you are regressing y on x and,
> say, y[3] is missing you can monitor node y[3] to see how Bugs is
> "imputing" it. Note that in order to impute an exogenous variable (x in
> this case). You would have to specify a distribution (presumably with
> unknown parameters) for x as well.
>
> You should NOT impute missing data on your own and run a Bugs program on
> it, let Bugs handle the missing values.
>
> ***********
> Best,
> Amalia
>
> -------------------------------------------------------------------
> This list is for discussion of modelling issues and the BUGS software.
> For help with crashes and error messages, first mail [log in to unmask]
> To mail the BUGS list, mail to [log in to unmask]
> Before mailing, please check the archive at
> www.jiscmail.ac.uk/lists/bugs.html
> Please do not mail attachments to the list.
> To leave the BUGS list, send LEAVE BUGS to [log in to unmask]
> If this fails, mail [log in to unmask], NOT the whole list
>
-------------------------------------------------------------------
This list is for discussion of modelling issues and the BUGS software.
For help with crashes and error messages, first mail [log in to unmask]
To mail the BUGS list, mail to [log in to unmask]
Before mailing, please check the archive at www.jiscmail.ac.uk/lists/bugs.html
Please do not mail attachments to the list.
To leave the BUGS list, send LEAVE BUGS to [log in to unmask]
If this fails, mail [log in to unmask], NOT the whole list
-------------------------------------------------------------------
This list is for discussion of modelling issues and the BUGS software.
For help with crashes and error messages, first mail [log in to unmask]
To mail the BUGS list, mail to [log in to unmask]
Before mailing, please check the archive at www.jiscmail.ac.uk/lists/bugs.html
Please do not mail attachments to the list.
To leave the BUGS list, send LEAVE BUGS to [log in to unmask]
If this fails, mail [log in to unmask], NOT the whole list
|