st: conformability issue associated with mi impute and clustered data

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

I am using Stata 12 via Windows 7.

I have annual data for a number of countries.  There are some missing
values for some variables.  I am trying to use -mi impute- to impute
these (clustered) missing values.

There is a very useful Stata FAQ about this
(http://www.stata.com/support/faqs/statistics/#mi).  It suggests that
one option is to include indicator variables for the clusters in the
imputation model.  If I follow this approach and I use -mi impute
regress- to impute the missing values for the two variables
separately, there is no problem.  The commands I use are:

. mi impute regress   NY_GNP_PCAP_CD            /*
    */    SP_DYN_LE00_IN     EN_POP_DNST     SP_RUR_TOTL_ZS    /*
    */    SP_RUR_TOTL_ZG     SP_URB_GROW     i.newcountryid, add(10) noisily

. mi impute regress    SE_PRE_ENRR_FE           /*
    */     SP_DYN_LE00_IN     EN_POP_DNST     SP_RUR_TOTL_ZS    /*
    */     SP_RUR_TOTL_ZG     SP_URB_GROW   i.newcountryid, add(10) noisily

But if I use a multivariate imputation strategy such as -mi impute mvn- and try

. mi impute mvn   NY_GNP_PCAP_CD    SE_PRE_ENRR_FE         /*
    */  =     SP_DYN_LE00_IN     EN_POP_DNST     SP_RUR_TOTL_ZS    /*
    */        SP_RUR_TOTL_ZG     SP_URB_GROW     i.newcountryid, add(10) force

then I get

Performing EM optimization:
note: 344 observations omitted from EM estimation because of all
imputation variables
      missing
conformability error
r(503);

Can anyone shed any light the 'conformability' error?

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/