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

Re: st: multiple imputation and instrumental variables and several tests

From   Maarten buis <>
Subject   Re: st: multiple imputation and instrumental variables and several tests
Date   Thu, 17 Dec 2009 10:45:43 -0800 (PST)

--- On Thu, 17/12/09, Matthijs De Zwaan <> wrote:
> I have used King et al's Amelia II program for use with R
> to create ten data sets that simulate the missing values in my panel
> data set. I would like to analyse these data in Stata 10.1 using
> -miest-. (My knowledge of R is rudimentary, and I do not have the time
> to learn R and King's Zelig plugin, although it seems that those
> programs would be able to give me what I want...)
> I want to do an Instrumental Variables analysis to deal
> with endogeneity issues, but -miest- is not compatible with
> -ivregress- or -ivreg2-. Is there a way to do IV with imputed data sets,
> preferably a straightforward one? If the re is no easy way, how should I
> go about this? Can I run my analysis for each of the ten data sets,
> and then combine the results in the same way that -miest- would
> combine a command, ie averages for the coefficients, and variance
> within data sets + variance across data sets for the standard errors?
> If so, can I do this for the first and second stages of TSLS separately?
> How should I deal with all the IV-diagnostics?

I don't know what -miest- is nor does -findit-. -findit- could find a 
reference to -mi_est-. This program has been superceded by -mim- 
(see: -findit mim-). This is again a reminder why it is crucial to 
specify where user written programs come from.

You'll have to specify the -category(fit)- option,  but otherwise it 
should work with -ivregress- and -ivreg2- (the latter is downloadable 
by typing -ssc install ivreg2-). IV diagnostics in combination with 
multiple imputation are probably very problematic, regardless of which 
program you use. So, if it has not been explicitly implemented and 
documented I would forget about those and not try to do that manually.

-- Maarten

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


*   For searches and help try:

© Copyright 1996–2017 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index