Statalist


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

RE: st: Restricted Ranges in Stata's Multiple Imputation Procedures


From   "Hoogendoorn, Adriaan" <A.Hoogendoorn@ggzingeest.nl>
To   "'statalist@hsphsun2.harvard.edu'" <statalist@hsphsun2.harvard.edu>
Subject   RE: st: Restricted Ranges in Stata's Multiple Imputation Procedures
Date   Tue, 20 Oct 2009 11:35:18 +0200

Dear Yulia,

Thank you for suggesting the knn option. I had not tried it before.
In my case it appears to reduce the "preasure" on the lowest observation. As mentioned before, the use of -mi impute pmm- with the default -knn(1)- option resulted in one imputed data set where the minimal observation (value 7) was matched to nine of the fourteen missings. Using -knn(3)- resulted into a reduced number of five of the fourteen missings that were matched to the minimal observation.

Thank you for considering an option to restrict the reanges of imputed values for -mi impute-.

Kind regards,
Adriaan Hoogendoorn
GGZ inGeest

-----Oorspronkelijk bericht-----
Van: owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu] Namens Yulia Marchenko, StataCorp LP
Verzonden: Monday, October 19, 2009 8:03 PM
Aan: statalist@hsphsun2.harvard.edu
Onderwerp: Re: st: Restricted Ranges in Stata's Multiple Imputation Procedures


Adriaan Hoogendoorn <A.Hoogendoorn@ggzingeest.nl> asks about restricting the range of imputed values to be within specified bounds during imputation using -mi impute-:

> I would like to restrict the range of imputed values and tried the
> (predictive mean) matching option. This technique resulted into an
> imputed data set where the minimal observation was (value 7) was
> matched to nine of the fourteen missings.

Adriaan uses the predictive mean matching imputation method to ensure that the imputed values are within the observed range.  By default, -mi impute pmm- replaces a missing value with the observed value for which the predicted value is the closest to that of the missing value.  In Adriaan's case, predicted values for most of missing observations were the closest to the minimum observed value.  To introduce more variation among the imputed values, Adriaan can increase the number of observations from which the imputed value is drawn by specifying the -knn()- option.  For example, using -knn(3)- with -mi impute
pmm- will result in randomly choosing a replacement value from 3 observations with closest predicted values.  There is no definitive recommendation in the literature on how many replacement values (nearest neighbors) should be used with predictive mean matching.  The choice depends on the bias/variance trade off and is specific to a particular application; see, for example, "[MI] mi impute pmm" for more detail.

> I would like to try a different solution. According to Allison (2001).
> "Some software can handle the restricted range problem in another way.
> If you specify a maximum or a minimum value for a particular variable,
> it will reject all random draws outside that range and simply take
> additional draws until it gets one within the specified range."(page
> 39). Am I correct that this option is not (yet) implemented in the ice
> package or the mi impute option?

Currently -mi impute- does not have an option for restricting the ranges of the imputed values during imputation.  We will consider adding this option in the future.


-- Yulia
ymarchenko@stata.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   |   What's new   |   Site index