[R] the output of function lars
Peter Ehlers
ehlers at ucalgary.ca
Thu Dec 9 01:17:45 CET 2010
I think the authors of lars could do a better job of
the 'Value' section of the help page. Nevertheless, most
of the list elements are pretty obvious if you
understand the LARS procedure. Your questions indicate
that you would benefit from studying the references
listed on the help page before attempting to use the
lars package. I can recommend the book 'The Elements
of Statistical Learning'.
Peter Ehlers
On 2010-12-08 08:02, poko2000 wrote:
>
> Hi here is the code as example
> lars is in package lars
>> x<-matrix(rnorm(20*5,0,1),20,5)
>> bs<-matrix(sample(seq(1:10),5),5,1)
>> er<-rnorm(20,0,1)
>> y<-x%*%bs+er
>> lobj<-lars(x,y,type="lasso")
>> names(lobj)
> [1] "call" "type" "df" "lambda" "R2"
> [6] "RSS" "Cp" "actions" "entry" "Gamrat"
> [11] "arc.length" "Gram" "beta" "mu" "normx"
> [16] "meanx"
>
> I do not know what does RSS, arc.length, mu mean???
> If mu means \hat{miu}, it does not seems to be right. I did not simulate my
> data with mean.
>
> besides I caculated R2 by myself but it cannot match the R2here
> I think R2=1-sum((y-x%*%lobj$beta)^2)/var(y)*19 #I have 20 observations...my
> R2 is lower than the one here
>
> Thanks a lot
>
>
>
More information about the R-help
mailing list