s-news
[Top] [All Lists]

Efficiency and robust regression

To: "S-new Listserve" <s-news@lists.biostat.wustl.edu>
Subject: Efficiency and robust regression
From: tangles@fastmail.fm
Date: Mon, 28 Feb 2005 06:59:13 -0800
Dear List,
After loading the robust library and running robust linear regression, I
notice that with some datasets I get indicators of bias like this:

Bias Tests for Robust Models:
Robust:
Test for Bias:
            Statistics  P-value 
 M-estimate      14.69 0.000645
LS-estimate       2.16 0.338916

When this happens I get an error statement that says: "final estimates
is not recommended. Use initial estimates as exploratory tools."

If I change the efficiency from the default of .9 to say, .7 I no longer
get indications of bias (efficiency is under the Advanced tab, in the
Final Estimates box). Can anyone tell me exactly what changing the
efficiency value does or refer me to an article or chapter that
discusses this issue? 
Thanks much,
Dan

<Prev in Thread] Current Thread [Next in Thread>
  • Efficiency and robust regression, tangles <=