<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta content="text/html;charset=ISO-8859-1" http-equiv="Content-Type">
</head>
<body bgcolor="#ffffff" text="#000000">
<tt><br>
Hi,<br>
The output of lmRob DOES provide the robust version of the R-squared
under “Multiple R-Squared”. It does not directly provide a test that
compares the model with the model with no covariates (equivalent of the
“F test”), but it can be obtained with the anova function as shown here:<br>
<br>
> creat.lmRob = lmRob(original1 ~
approprie1+approprie2+creativite1+creativite2, data=creatif)<br>
> creat00.lmRob = lmRob(original1 ~ 1, data=creatif)<br>
> anova (creat00.lmRob, creat.lmRob)<br>
<br>
This can also be done with lmrob().<br>
<br>
Concerning the R-squared however, we have recently published a paper
that shows that the robust R-squared provided by lmRob is biased,
sometimes to a large extent. We provide a consistent and robust
estimator of R-squared (robR2w.WithCorrection in the output below) and
a version adjusted for the sample size (robR2w.AdjustedWithCorrection
in the output below). The code and an example are provided below.<br>
<br>
Olivier<br>
<br>
ref: Renaud, O. & Victoria-Feser, M.-P. (2010). A robust
coefficient of determination for regression. Journal of Statistical
Planning and Inference, 140, 1852-1862.
<a class="moz-txt-link-freetext" href="http://dx.doi.org/10.1016/j.jspi.2010.01.008">http://dx.doi.org/10.1016/j.jspi.2010.01.008</a><br>
<br>
> library(robust)<br>
> source("robR2w.r")<br>
<br>
> creat.lmRob = lmRob(original1 ~
approprie1+approprie2+creativite1+creativite2, data=creatif)<br>
<br>
> summary(creat.lmRob)<br>
<br>
Call: lmRob(formula = original1 ~ approprie1 + approprie2 + creativite1
+ <br>
creativite2, data = creatif)<br>
<br>
Residuals:<br>
Min 1Q Median 3Q Max <br>
-1.96149388 -0.34543174 -0.05500626 0.23168813 1.73781067 <br>
<br>
Coefficients:<br>
Value Std. Error t value Pr(>|t|) <br>
(Intercept) -2.2744444543 1.2656121996 -1.7971100903 0.0784825453<br>
approprie1 0.0914187017 0.1187669959 0.7697315322 0.4451541682<br>
approprie2 0.1505246740 0.0934840433 1.6101643524 0.1137861853<br>
creativite1 0.6270578015 0.1648705921 3.8033332300 0.0003963918<br>
creativite2 -0.2384886952 0.1302348886 -1.8312197120 0.0731512215<br>
<br>
Residual standard error: 0.459508 on 49 degrees of freedom<br>
Multiple R-Squared: 0.183627 <br>
<br>
Test for Bias:<br>
statistic p-value<br>
M-estimate 6.527018 0.25825825<br>
LS-estimate 10.368150 0.06545114<br>
<br>
> robR2w(creat.lmRob)<br>
$robR2w.NoCorrection<br>
[1] 0.3732078<br>
<br>
$robR2w.WithCorrection<br>
[1] 0.3302368<br>
<br>
$robR2w.AdjustedWithCorrection<br>
[1] 0.2604698<br>
<br>
<br>
<br>
friendpine<big> </big>wrote:</tt>
<blockquote cite="mid:loom.20100324T044117-751@post.gmane.org"
type="cite">
<pre wrap=""><tt>The output for the rlm(),lmRob(),lmrob() didn't give the R-squared and the
p-value of the equation. There seems to be no function that can do these
things.Can we calculate it as in the OLS(ordinary least squared) regression? The
method in OLS use the SSE and SSR to calculate the f for F test. The p-value
generated in the F test is the p-value for the OLS regression equation. Can we
do it in the robust regression?
_______________________________________________
<a class="moz-txt-link-abbreviated" href="mailto:R-SIG-Robust@r-project.org">R-SIG-Robust@r-project.org</a> mailing list
<a class="moz-txt-link-freetext" href="https://stat.ethz.ch/mailman/listinfo/r-sig-robust">https://stat.ethz.ch/mailman/listinfo/r-sig-robust</a>
</tt></pre>
</blockquote>
<tt><br>
</tt>
<pre class="moz-signature" cols="72"><tt>--
<a class="moz-txt-link-abbreviated" href="mailto:Olivier.Renaud@unige.ch">Olivier.Renaud@unige.ch</a> <a class="moz-txt-link-freetext" href="http://www.unige.ch/fapse/mad/">http://www.unige.ch/fapse/mad/</a>
Methodology & Data Analysis - Psychology Dept - University of Geneva
UniMail, Office 4164 - 40, Bd du Pont d'Arve - CH-1211 Geneva 4
</tt></pre>
</body>
</html>