Re: st: mi with imputed data sets only?

[Steven Samuels <sjsamuels@gmail.com>]

Follow-up:

It was easy to recreate the original (m = 0) data set. To do so, I calculated for each respondent the SD over imputations of each of the possibly imputed variables.  The SD would be zero for values that had not been imputed and >0 for imputed observations.  

I took one imputed set, and set each possibly imputed variable  to missing if the SD of that variable over imputations was >0, changed the imputation indicator to 0 and appended it to the original data. -foreach- loops over the list of possibly imputed variables helped a lot.

Interestingly, the SD analysis found about 80 observations that are erroneously coded as having no imputations.

Anyone who would like to see the code for these analyses is welcome to write to me privately.

Steve
sjsamuels@gmail.com


On Jul 19, 2011, at 9:11 AM, Steven Samuels wrote:


The US National Health and Examination Survey (NHANES) distributes multiply imputed data that consist of five imputation data sets.  These are The 1999-2004 Dual Energy X-ray Absorptiometry (DXA) Multiple Imputation Data Files (http://www.cdc.gov/nchs/nhanes/dxx/dxa.htm). I cannot use Stata's -mi import- commands, because theserequire the original (m = 0) data.

I'm going to try to recreate the m = 0 data  from the supplied imputation indicators.  But before I do, I'd like to know:  Is it possible to use -mi estimate- with only the imputed data sets?  If so, how?

Thanks,

Steve

sjsamuels@gmail.com






*
*   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/


*
*   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/