Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, statalist.org is already up and running.


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

Re: st: Multiple imputation: negative imputed values.


From   Alan Acock <acock@mac.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: Multiple imputation: negative imputed values.
Date   Thu, 30 Dec 2010 10:07:37 -0800

The negative imputed values are the values consistent with your observed data. You might think of them as place holders letting you use all your observed data rather than dropping observations with missing values. Some people change negative values to zeros or the lowest observed value on a variable, but this is adding information to your data rather than using all available information in your data. If there are a large number of out of bound values, you probably have trouble with the missing at random assumption. You might benefit by adding auxiliary variables that predict missing values (that don't have a lot of missing values themselves).

It is possible to eliminate negative values and the ice command using chained equations does this nicely, but some would argue that in so doing you are adding information to the dataset rather than preserving all of the information in your dataset.

==Alan Acock

On Dec 30, 2010, at Thu Dec 5 7:29 , Khoi Dinh To wrote:

> I was working on a data set with a lot of missing data.  I tried using multiple imputation and the results looked better.  However, I have 2 concerns as follows:
> 
> 1. A few original variables do not have negative values (for example, faculty salaries, the number of bachelor's degrees produced...).  However, after I did 5 imputations (5 iterations), I found out that there were negative imputed values in those variables, too.  Are those negative values acceptable (because the original data went through 5 iterations)?  If they are not, then is there any measure to correct them?  I am asking these questions because I think I will have to present the descriptives of the imputed data, and I want to make sure the negative imputed values are acceptable to the public.
> 
> 2. After the imputation, I ran the command sureg (seemingly unrelated regression).  However, Stata did not report the adjusted R-squared or RMSE.  Is there any way to get these statistics from the output?
> 
> Thank you very much.
> 
> ----------------------------------
> Khoi Dinh To
> E-mail: todinhkhoi@yahoo.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/


© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index