[R] FW: AIC Quantile Regression

frank frank beth_strain2002 at hotmail.com
Wed Feb 13 01:39:28 CET 2008




> To: beth_strain2002 at hotmail.com
> From: estrain at postoffice.utas.edu.au
> Date: Wed, 13 Feb 2008 11:38:43 +1100
> Subject: AIC Quantile Regression
> 
> Dear R,
> I currently trying to fit quantile regression models to test the
> relationship between abalone and algal cover at different quadrat sizes.
> I am using the AIC values as a measure of model fit. Each model has the
> same number of parameters and data points but uses different data sets,
> for example
> 
> fit1<-rg(log(Hrubra+0.001)~%coverSessileInvertebrates, tau=0.9,
> data=Quadrat0.25 x 0.25m)
> fit1<-rg(log(Hrubra+0.001)~%coverSessileInvertebrates, tau=0.9,
> data=Quadrat0.5 x0.5m)
> etc
> An example of the data is below. After look through the help files I
> still have a couple of questions. My questions are 
> 
> 1. Is this an appropriate use of AIC given that I am testing different
> data sets?
> 2. Why does is the AIC going down in value although the fit of the models 
> appear to be getting worse as the size of the model increases
> 3. How are the AIC values calculated in quantile regression.
> 
> Thanks in advance
> Beth Strain
> 
> 
> 
> 
> 
> 
> 
> 
> 
> 

_________________________________________________________________
[[elided Hotmail spam]]

eferral=hotmailtaglineOct07&URL=http://music.ninemsn.com.au/roguetraders


More information about the R-help mailing list