[R] What goodness-of-fit measure for robust regression ?

[Liaw, Andy]

This is at best on the order of a match light...

GOF and robust estimation seem to be answering rather conflicting questions.
GOF asks how well the model fit the data.  Robust methods say not to pay too
much attention to some "bad" data points.  How does one tell GOF which
subset of data to pay less attention to?

One possible question is that, with respect to the data that the robust
estimation procedure treated as "good", how good a fit is the linear model?
I do not know of any work on this.

Andy

> From: vincent.vicaire at libertysurf.fr 
> 
> Hi,
> 
> 
> i. After estimating some coefficients using robust regression 
> with rlm() or lqs(), I wonder if there exist some measures of 
> the goodness-of-fit as those for standard linear model(R2)... 
> or evenly if it's a statistics non-sense to look for since I 
> do not find any mention of that in differents chapters on 
> robust and resistant regression or in severals R 
> documentation (Fox, Ripley and Venables)?
> 
> ii.If such measure exist for robust regression, how could I 
> implement it in R, and what about their comparability with 
> traditionnal R2 of linear model ?
> 
> Best regards
> A lot of thank to anyone that could give me some light.
> 
> Vincent
> 
> ********** L'ADSL A 20 EUR/MOIS********** 
> Avec Tiscali, l'ADSL est à 20 EUR/mois. Vous pourrez chercher 
> longtemps avant de trouver moins cher ! 
> Pour profiter de cette offre exceptionnelle, cliquez ici : 
http://register.tiscali.fr/adsl/ Offre soumise à conditions.

______________________________________________
R-help at stat.math.ethz.ch mailing list
https://www.stat.math.ethz.ch/mailman/listinfo/r-help