[S] lo() within gam() vs. loess()

Steven Paul Millard (probstat@nwrain.com)
Mon, 9 Mar 1998 15:41:01 -0800


Hello,

I posted this question about a week and a half ago but have not
received any replies, so I'll try one more time.

Can someone please explain to me why the following fits of lo() called
within gam() vs. loess() are not the same? (Ah, but Dr. Ripley will
probably want me to say "Will someone please explain ....")

Also, is there a theoretical reason why you can't tell lo() to assume
family="symmetric" instead of family="gaussian"? (I guess this option
is really taken care of if you use the robust() function when
specifying the family argument.)

Thank you for your time!!

> dum.x <- seq(0, 2*pi, len=100)
> set.seed(10)
> dum.y <- sin(dum.x) + rnorm(100)
> plot(dum.x, dum.y)
> fit.gam <- gam(dum.y ~ lo(dum.x, span=0.5, degree=1))
> fit.loess <- loess(dum.y ~ dum.x, span=0.5, degree=1)
> summary(fit.gam$fitted - fit.loess$fitted)
Min. 1st Qu. Median Mean 3rd Qu. Max.
-0.0482 -0.01954 0.009125 0.009125 0.03779 0.06646

Sincerely,

--Steve M.

_____________
| *** | Steven P. Millard, Ph.D.
| * |
| * * * | P robability, TEL: 206-528-4877
| * * * | S tatistics & FAX: 206-528-4802
| * | I nformation E-mail: SMillard@ProbStatInfo.com
| * | Web: www.ProbStatInfo.com
| *** | 7723 44th Avenue NE
|___________| Seattle, WA 98115-5117 USA

-----------------------------------------------------------------------
This message was distributed by s-news@wubios.wustl.edu. To unsubscribe
send e-mail to s-news-request@wubios.wustl.edu with the BODY of the
message: unsubscribe s-news