Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: mahapick updates


From   David Kantor <kantor.d@att.net>
To   Statalist <statalist@hsphsun2.harvard.edu>
Subject   st: mahapick updates
Date   Thu, 03 Apr 2008 11:52:54 -0400

Thanks to Kit Baum, updates to the mahapick package are now available on SSC.

The mahapick package has programs to calculate a distance measure (propensity score, dissimilarity measure) between observations, and to choose a set of close matches for a selected subset of observations.

There are two important changes:

1, this now implements a true Mahalanobis distance measure -- whereas previously, it was the normalized Euclidean distance. The true Mahalanobis measure accounts for correlation between variables. (The old behavior is available as an option.)

2, the new program mahascores enables you to directly obtain the distance measures between all pairs of observations. The concept of "treated" observations is optional. And, while there is some overlap with some of the functionality of mahapick, this would be the preferred method if all you want is the distance measure (and don't need, at this stage, to select the closest matches). The results can be placed into the current dataset, a separate dataset, or a matrix.

There are other enhancements that should give users more control.

Thanks also go to Heiko Giebler of Wissenschaftszentrum Berlin fur Sozialforschung gGmbH -- for the suggestions and advice that led to these improvements.

To download...
ssc desc mahapick
ssc inst mahapick

If any users try it and have any problems, comments, or suggestions, I'd like to hear them.
Best wishes,
--David Kantor

*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* 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