R help - Jun 2005 - mulitple mixed regression with spline

Hi, I'm looking to implement a regression with mixed terms. I have 2
biological endpoints for a dataset of n=77, one linearly related and
the other fits a spline.  I want to combine these two terms in a
linear regression for prediction, then apply the model to a test set.

this works fine, good r2 and I've graphed the spline.
m1<-lm(y~x1,data=train)
m2<-smooth.spline(x2,y); (spl)

what i want is
y=x+bilin(x2)

any tips? sorry, fairly new user.

J.

On Fri, 24 Jun 2005, james lumley wrote:
> Hi, I'm looking to implement a regression with mixed terms. I have 2
What do you mean by `mixed'?  Not I think  in the sense of Pinheiro & 
Bates' book or nlme.
> biological endpoints for a dataset of n=77, one linearly related and
> the other fits a spline.  I want to combine these two terms in a
> linear regression for prediction, then apply the model to a test set.
>
> this works fine, good r2 and I've graphed the spline.
> m1<-lm(y~x1,data=train)
> m2<-smooth.spline(x2,y); (spl)
>
> what i want is
> y=x+bilin(x2)
You can see several ways to do this in MASS.  Most simply

lm(y ~ x1 + ns(s2, df=?))

for regression splines.  For smoothing splines, see the functions gam() in
packages mgcv and gam (which differ considerably), or bruto() in mda or 
package gss or ....

-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595