Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, statalist.org is already up and running.


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: st: mi impute chained


From   chong shiauyun <shiauyun416@hotmail.com>
To   <statalist@hsphsun2.harvard.edu>
Subject   RE: st: mi impute chained
Date   Wed, 31 Oct 2012 16:54:50 +0800

Hi, 

thanks for your advice.
I simplified my MI model by excluding some interactions and reduced some of my variables. It works fine. However, I am concern that I have to use the -force- option  to make the model works. It am quiet reluctant to drop all of the interactions seeing that it may affect the relationship between the exposure and the outcome which I am interested in.

Kind regards
Shiau

----------------------------------------
> Date: Fri, 26 Oct 2012 08:20:35 -0400
> Subject: Re: st: mi impute chained
> From: jvverkuilen@gmail.com
> To: statalist@hsphsun2.harvard.edu
>
> As I said before, your model is way too complex to troubleshoot. Cut
> down on the number of variables or simplify the imputation model and
> build UP. There are simply too many failure points right now to be
> able to diagnose anything. For instance I seem to recall you're using
> a multinomial logit for a categorical variable. You may have problems
> there. The censored parts may be trouble too. With imputation the goal
> is to get a "good enough" representation of the dataset. So cut down
> to a much simpler model. See if that runs. Then start adding
> components and find out where it fails. Try this a few different ways.
> If it always fails on the same variable that tells you what's wrong.
> I'd also run descriptives and missing data patterns on all variables.
>
>
>
> On Fri, Oct 26, 2012 at 3:44 AM, chong shiauyun <shiauyun416@hotmail.com> wrote:
> > Hi,
> >
> > I am not sure if I should proceed to imputation or I need to change my dryrun model because in my dryrun, it seems to have some convergence problems. I am not sure how to check which sub-model that caused the convergence problem, even though I have specified -noisily-. The thing is that the iterations keep running something like:
> > Iteration 611: log likelihood = -12162.404 (not concave)
> > Iteration 612: log likelihood = -12162.4 (not concave)
> > Iteration 613: log likelihood = -12162.397 (not concave)
> > Iteration 614: log likelihood = -12162.394 (not concave)
> > Iteration 615: log likelihood = -12162.39 (not concave)
> > Iteration 616: log likelihood = -12162.387 (not concave)
> > Iteration 617: log likelihood = -12162.384 (not concave)
> > Iteration 618: log likelihood = -12162.38 (not concave)
> >
> > Shiau
> >
> > ----------------------------------------
> >> From: ymarchenko@stata.com
> >> To: statalist@hsphsun2.harvard.edu
> >> Subject: Re: st: mi impute chained
> >> Date: Thu, 25 Oct 2012 07:21:35 -0500
> >>
> >> Chong Shiauyun <shiauyun416@hotmail.com> receives a "no observation" error
> >> when he runs the following imputation model using -mi impute chained-:
> >>
> >> > . mi impute chained (reg) birthweight (ologit, augment) ednmatpat
> >> > (logit, augment)sex (truncreg, ll(lVIQ) ul(uVIQ))verbiq,
> >> > add(20) rseed(11349730) burnin(50) chainonly dryrun report
> >>
> >> Shiau probably meant to omit the -dryrun- option in the above since
> >> -mi impute chained- does not perform any estimation when the -dryrun- option
> >> is specified.
> >>
> >> The -mi impute chained- command starts off by fitting univariate models on the
> >> observed data to obtain initial imputed values for each imputed variable. The
> >> "no observation" error typically occurs when one of such models contains no
> >> observations and is often caused by the existence of missing values in
> >> variables other than the imputed variables. In Shiau's case, the offending
> >> variables may be -lVIQ- and -uVIQ-.
> >>
> >> Shiau can use the -noisily- option of -mi impute chained- to identify the
> >> particular conditional model for which there are no observations and run this
> >> model manually on the observed data to determine the problem.
> >>
> >> Shiau can also send his data and do file to our technical support group at
> >> tech-support@stata.com to help him identify the problem.
> >>
> >>
> >> -- Yulia
> >> ymarchenko@stata.com
> >> *
> >> * For searches and help try:
> >> * http://www.stata.com/help.cgi?search
> >> * http://www.stata.com/support/faqs/resources/statalist-faq/
> >> * http://www.ats.ucla.edu/stat/stata/
> >
> > *
> > * For searches and help try:
> > * http://www.stata.com/help.cgi?search
> > * http://www.stata.com/support/faqs/resources/statalist-faq/
> > * http://www.ats.ucla.edu/stat/stata/
>
>
>
> --
> JVVerkuilen, PhD
> jvverkuilen@gmail.com
>
> "Out beyond ideas of wrong-doing and right-doing there is a field.
> I'll meet you there. When the soul lies down in that grass the world
> is too full to talk about." ---Rumi
>
> *
> * For searches and help try:
> * http://www.stata.com/help.cgi?search
> * http://www.stata.com/support/faqs/resources/statalist-faq/
> * http://www.ats.ucla.edu/stat/stata/
 		 	   		  
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/


© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index