st: multiple imputation with a panel data set in Stata 12

[Stephen Martin <stephen.martin@york.ac.uk>]

Hello Statlisters,

I have a panel data set (40 countries with 30 annual observations
on, say, 50 variables).  The dependent variable, y, has no missing
values.  Five regressors (z1 - z5) have no missing values but 44 other
potential regressors (x1 - x44) have varying degrees of 'missingness'.

As I understand it, the first two steps with the use of imputed data
involve (a) the generation of the imputed data sets and (b) the
analysis of these imputed data sets.

It is clear to me how to undertake step (b) because the -mi estimate-
command in Stata 12 supports the use of xtreg to estimate a panel
model across some already imputed data sets.

However, it is unclear to me how I should construct the necessary
imputed datasets (i.e., undertake step (a)).  For example, the command
-mi impute chained- supports a variety of regression methods but none
of these would appear to respect the panel nature of the data.  Am I
missing or misunderstanding something?

Any help or advice would be most welcome.

Thanks.

Steve
*
*   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/