Re: st: mi impute: ologit, mlogit, logit in one equation?

[Maria Fleischmann <maria.s.fleischmann@gmail.com>]

Thanks, Richard, for this suggestion. I will consider it!

But just out of interest: is it possible to impute missing values with
a non-monotone missing structure if the variables should be imputed
with different regression equations (mlogit, logit, reg)?
Which should be something like:
mi impute (regress) var1 (ologit) var2 var3 (mlogit) var5 = var0 var4
i.var6 var7 i.var8, replace



On 17 May 2011 15:49, Richard Goldstein <richgold@ix.netcom.com> wrote:
>
> Maria,
>
> since you have already imputed the data, why not use -mi import- and
> then you can use -mi estimate- (or any other built-in tool)?
>
> Rich
>
> On 5/17/11 9:43 AM, Maria Fleischmann wrote:
> > Dear statalister,
> >
> > I am currently trying to impute multiple individual-level variables
> > from a clustered dataset.
> > I did so with stata command ice, but I would rather like to use mi
> > impute due to the mi estimate command.
> > My equation in ice (without taking into account the layered structure
> > of the data) was the following, where I impute variabels by ologit,
> > mlogit, logistic or linear regression, all delineated in one command:
> >
> > ice var0 var1 o.var2 o.var3 var4 m.var5 i.var6 var7 i.var8, m(5)
> > saving(20110511_ice_1, replace)
> >
> > Now I would like to transfer this to a mi impute command, but this
> > raises the question whether I can use mlogit, ologit, linear
> > regression in the same equation if the missing structure in my data is
> > not monotonic?
> > I think the mi impute command should look somehow like this:
> >
> > mi set mlong
> > mi set M=5
> > mi register imputed var1 var2 var3 var5
> > mi register regular var0 var4 var6 var7 var8
> > mi impute (regress) var1 (ologit) var2 var3 (mlogit) var5 = var0 var4
> > i.var6 var7 i.var8, replace
> >
> > This however does not work, because the data is not monotonic?
> > So I tried to do the imputation separate for all the different
> > regression types, which also does not work, because the missing
> > structure is still not monotonic and I have more variables for one
> > type of regression, e.g. ologit for var2 and var3.
> > But also when doing the imputation separate for each variable (which I
> > find highly inconvenient especially due to the large number of
> > variables I want to impute), stata reports problems because I am
> > imputing variables with variables that also have missing values. Do I
> > just have to force stata to impute or is there a more elegant way of
> > solving my problem (which I hope and think there is)?
> > These are all problems I encountered besides the fact that I also
> > would like to account for the clustering of the data.
> >
> > Thanks you for your help!
> >
> > Maria
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/