Overfitting in snr (ASSIST)

Previous Topic Next Topic
 
classic Classic list List threaded Threaded
1 message Options
Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Overfitting in snr (ASSIST)

Simon Frost
Dear R-Help group,

I'm trying to fit some smoothing spline models using the snr function in the
ASSIST library. My model is the following:

y~log(exp(a0-a1*x)+exp(b0-b1(x)))

where b1(x) is a nonparametric function. It's basically a biexponential decay,
with wiggliness in 'phase 2' decay.

I can easily fit this using regression splines, but I'd like to see how
smoothing splines work in this scenario. I tried to use the snr function, as
follows

snr(y~log(exp(a0-a1*x)+exp(b0-b1(x))),func=list(b1(u)~list(~u,cubic2
(u))),params=list(a0+a1+b0~1),data=mydata,start=list(params=mystart))

This runs happily, but interpolates the data, rather than smoothing it. This
is probably due to some identifiability issues, but I'm not sure how to
parameterize the model to fix this. Any suggestions?

Best wishes
Simon

______________________________________________
[hidden email] mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
Loading...