R help - Apr 2004 - predict( ..., type="terms")

Hello,

After reading the help for predict.lm and predict.glm, it is not clear
to me what are the values returned by predict( ..., type="terms"). 

Anybody willing to enlighten me?

The example provided by Peter Dalgaard in a recent post unfortunately
was enlightening only to the point of making me realize that "terms"
are not what I thought they were.

Thanks in advance,
Giovanni

-- 

 __________________________________________________
[                                                  ]
[ Giovanni Petris                 GPetris at uark.edu ]
[ Department of Mathematical Sciences              ]
[ University of Arkansas - Fayetteville, AR 72701  ]
[ Ph: (479) 575-6324, 575-8630 (fax)               ]
[ http://definetti.uark.edu/~gpetris/              ]
[__________________________________________________]

On Wed, 7 Apr 2004, Giovanni Petris wrote:
>
> Hello,
>
> After reading the help for predict.lm and predict.glm, it is not clear
> to me what are the values returned by predict( ...,
type="terms").
>
> Anybody willing to enlighten me?
>
For each term in the formula, extract its coefficients and the
corresponding columns of the design matrix, and multiply the two, and then
center the result.

So, looking at the example in example(termplot) we have a model
y~ns(x,6)+z where z is a factor with 4 levels.

There are 9 coefficients in the model, but only two terms: ns(x,6) (based
on 5 coefficients) and z (based on 3 coefficients).

Adding up the terms and the centering value (also returned by
predict(,type="terms")) gives the linear predictor.

	-thomas

> -----Original Message-----
> From: r-help-bounces at stat.math.ethz.ch 
> [mailto:r-help-bounces at stat.math.ethz.ch] On Behalf Of Thomas Lumley
> Sent: Wednesday, April 07, 2004 5:44 PM
> To: Giovanni Petris
> On Wed, 7 Apr 2004, Giovanni Petris wrote:
> 
> >
> > Hello,
> >
> > After reading the help for predict.lm and predict.glm, it 
> is not clear 
> > to me what are the values returned by predict( ...,
type="terms").
> >
> > Anybody willing to enlighten me?
> >
> 
> For each term in the formula, extract its coefficients and 
> the corresponding columns of the design matrix, and multiply 
> the two, and then center the result.
> 
> So, looking at the example in example(termplot) we have a 
> model y~ns(x,6)+z where z is a factor with 4 levels.
> 
> There are 9 coefficients in the model, but only two terms: 
> ns(x,6) (based on 5 coefficients) and z (based on 3 coefficients).
> 
> Adding up the terms and the centering value (also returned by
> predict(,type="terms")) gives the linear predictor.
1) It would be useful to have this in the help file for predict.(g)lm
   Also that the constant is accessed as attr( pr.obj, "const" )

2) A link to termplot might be useful too.

3) I guess that Giovanni (as well as I) was curious what the rationale
behind
   the centering of effects is. Can anyone tell us?

Bendix Carstensen

----------------------
Bendix Carstensen
Senior Statistician
Steno Diabetes Center
Niels Steensens Vej 2
DK-2820 Gentofte
Denmark
tel: +45 44 43 87 38
mob: +45 30 75 87 38
fax: +45 44 43 07 06
bxc at steno.dk
www.biostat.ku.dk/~bxc