On Sun, 23 Mar 2003, Davorka Gulisija wrote:
> I am using Loess procedure in order to infer certain relationship
> for the data set of ~130,000 observations with ~300 distinct values of
> single predictor. I understand that fitted values y-hat are just 300
> Weighted
> LS fits in certain neighborhood of predictors. I am bit confused about
> how exactly is this neighborhood assigned . Say I choose spanning
> parameter = .5, for each LS analysis 75000 observations should be used.
> However, intuitively it doesn't seem right since points are not equally
> distributed among predictors and there are many observations for a single
> value of predictor.
> I would appreciate if someone could clear this for me.
I have an interactive demo to help understand what is going on with the
loess function. You can download a copy of the demo by going to my
website at: http://statistics.byu.edu/faculty/faculty.php?person_id=34 and
click on the link under programming and scrolling down to near the bottom
and clicking on the loess.demo.ssc link.
You run the function with a command like:
attach(ethanol)
loess.demo(E,NOx,span=.5, degree=1)
and it will plot the data along with the loess line (actually 2, the extra
smoothed line and the raw loess line), clicking on the graph will then
show the window around the x value you clicked, the weights assigned to
the points within that window, and the fitted line used to predict y for
the clicked x value. Click again to see what happens with a differnt x
value. Right click to end the demo.
Hope this helps you understand the loess function better,
--
Greg Snow, PhD Office: 223A TMCB
Department of Statistics Phone: (801) 378-7049
Brigham Young University Dept.: (801) 378-4505
Provo, UT 84602 email: gls@byu.edu
|