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

Re: st: Postestimation procedures multiple data sets

From   Maarten buis <>
Subject   Re: st: Postestimation procedures multiple data sets
Date   Mon, 27 Apr 2009 08:50:37 +0000 (GMT)

--- On Mon, 27/4/09, wrote:
> I performed an ice command that generated 5-data sets (each
> n = 3,300). I followed up with a mim run (regress). I am
> using Stata 10. I read the ice and mim command help files.
> It appears to me that I cannot obtain an R^2 value for the
> overall mim run. It also seems to me that I cannot perform
> any multivariate diagnostics for assumption violations on
> that run of the data. Is this correct? 

I would limit the model checking to graphical checks (I 
would do so anyhow even if I used non-imputed data, but 
that is another issue and opions differ on this). An 
example of how you could do that is shown below:
*---------- begin example ----------------
sysuse auto, clear
gen miss = missing(rep78, mpg, price)
ice rep78 mpg price, clear m(5) 

mim, storebv: reg price mpg rep78
predict xb
gen resid = xb -  price

twoway scatter resid mpg             ///
       if miss == 0 ,                ///
       by(_mj, compact) ||           ///
       scatter resid mpg             ///
       if miss == 1,                 ///
       mcolor(red)                   ///
       legend(order(1 "observed"     ///
                    2 "imputed"))    ///
       ytitle(residual) ||           ///
       mband resid mpg
*-------------- end example -----------------	

As to R^2, see the opion of Don Rubin here:

and my interpretation of it in Stata code here:

Hope this helps,

Maarten L. Buis
Institut fuer Soziologie
Universitaet Tuebingen
Wilhelmstrasse 36
72074 Tuebingen


*   For searches and help try:

© Copyright 1996–2015 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index