Statalist The Stata Listserver


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

Re: st: getting rid of the outliners


From   Ronán Conroy <rconroy@rcsi.ie>
To   "statalist hsphsun2.harvard.edu" <statalist@hsphsun2.harvard.edu>
Subject   Re: st: getting rid of the outliners
Date   Wed, 3 May 2006 15:25:46 +0100

On 30 Aib 2006, at 01:47, vora n wrote:

Is there any STATA command that can drop
the observations that are the outliners?

Dropping observations is never a good idea. If we restrict reasoning to the data that fit our model, we end up with something more like CNN than science.

If your data have outliers, then a couple of checks are needed:

1. Are they real?
If they are simply errors, you can rectify them

2. Are they all one one side of the distribution?
If so, you may have a distribution that is better re-expressed on a different scale, such as log-transformed.

3. Are they causing you to produce misleading conclusions?
Here, there are a couple of approaches possible in Stata:
-rreg- performs weighted regression which does not eliminate your outliers, but downweights their influence on the regression. Check to see that associations which were significant do not change to non- significant and vica versa.

-ologit- Try re-expressing your outcome variable in, say, deciles, and using -ologit- to model it. If the conclusions differ from the analysis you already performed, distrust the latter. You can also use -ologit-, of course, on your original data, so long as you don't have too many individual values - the limit is 50.





=========
Ronán Conroy
rconroy@rcsi.ie
+353 (0) 1 402 2431
+353 (0) 87 799 97 95
www.flickr.com/photos/ronanconroy




*
* 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