Skip to content

plotting gam models

7 messages · Lucia Zarauz, Henrique Dallazuanna, Rod +1 more

Original
Lucia Zarauz
# 
Hi everybody,

I am working with gams and I have found some questions when plotting gams models.

I am using mgcv, and my model looks something like this:

model<- gam(x ~ s(lat,long))

I can plot the output of the model using plot(model) or plot.gam(model) and I get a surface plot. 

That is ok, but what I want to do now is to extract the data used to perform the surface plot. Like that I would be able to superpose them to a geographical map, and to plot these data using other programs.

Thank you very much in advance

Luc?a
_____________________________________________ 
Lucia Zarauz 
AZTI - Tecnalia / Unidad de Investigaci?n Marina
Herrera kaia portualdea z/g
20110 Pasaia (Gipuzkoa)
Tel: 943 004 800 - Fax: 943 004 801
e-mail: lzarauz at pas.azti.es
www.azti.es ; www.tecnalia.info
_____________________________________________ 
? 
**************** LEGE OHARRA **************** AVISO LEGAL **************** DISCLAIMER ****************
Mezu hau pertsonala eta isilpekoa da eta baimenik gabeko erabilera debekatua dago legalki. Jasotzailea ez bazara ezabatu mezua, bidali eta kontserbatu gabe. 
Este mensaje es personal y confidencial y su uso no autorizado est? prohibido legalmente. Si usted no es el destinatario, proceda a borrarlo, sin reenviarlo ni conservarlo.
This message is personal and confidential, unauthorised use is legally prohibited. If you are not the intended recipient, delete it without resending or backing it.
Lucia Zarauz
# 
Hi Henrique,

Thank you for your suggestion. 

Actually, I have already tried it, but I am confused because the plot I get is not the same as the output of plot(model) or plot.gam(model). The yaxis is different

On the other hand, if I build a more complex model, as for example:

model<- gam(x ~ s(lat,long) + s(temperature))

I would like to extract the information to build the effects for each explanatory factor (one graph for s(lat,long) and another for s(temperature)), as R does when you use 'plot(model)' and you press return for subsequent pages.

My final aim is to plot the influence of s(lat,long) as a contourplot superposed on a geographical map. Maybe there is an easier way to do it...

Thank you very much

luc?a



_____________________________________________ 
Lucia Zarauz 
AZTI - Tecnalia / Unidad de Investigaci?n Marina
Herrera kaia portualdea z/g
20110 Pasaia (Gipuzkoa)
Tel: 943 004 800 - Fax: 943 004 801
e-mail: lzarauz at pas.azti.es
www.azti.es ; www.tecnalia.info
_____________________________________________ 
? 
**************** LEGE OHARRA **************** AVISO LEGAL **************** DISCLAIMER ****************
Mezu hau pertsonala eta isilpekoa da eta baimenik gabeko erabilera debekatua dago legalki. Jasotzailea ez bazara ezabatu mezua, bidali eta kontserbatu gabe. 
Este mensaje es personal y confidencial y su uso no autorizado est? prohibido legalmente. Si usted no es el destinatario, proceda a borrarlo, sin reenviarlo ni conservarlo.
This message is personal and confidential, unauthorised use is legally prohibited. If you are not the intended recipient, delete it without resending or backing it.
Simon Wood
# 
-- `plot.gam' will always plot the `centered smooth', i.e.  the smooth 
constrained to sum to zero over the data. By default (i.e. using type="link") 
`predict.gam' will include the intercept in the predctions. If you want 
centered terms out of `predict.gam' use the `type="terms" option.
Again, try predict.gam with type="terms"

best,
Simon
Rod
# 
2007/7/24, Lucia Zarauz <lzarauz at pas.azti.es>:
Also see ?vis.gam.
Produces perspective or contour plot views of gam model predictions.

Rod.
Simon Wood
# 
What's wrong with the predict.gam(...,type="terms") suggestion (see 
yesterday)?

Here's example code:
## example data from ?gam
n<-400;sig<-2
x <- runif(n, 0, 1);z <- runif(n, 0, 1)
x2 <- runif(n, 0, 1);x3 <- runif(n, 0, 1)
f0 <- function(x) 2 * sin(pi * x)
f1 <- function(x) exp(2 * x)
f2 <- function(x) 0.2*x^11*(10*(1-x))^6+10*(10*x)^3*(1-x)^10
f3 <- function(x) 0*x
f <- f0(x) + f1(z) + f2(x2)
e <- rnorm(n, 0, sig)
y <- f + e
b<-gam(y~s(x,z)+s(x2)+s(x3))

## make a grid of data over which to predict
zm <- xm <- seq(0,1,length=40)
pd <- data.frame(x=rep(xm,40),z=rep(zm,rep(40,40)),
                 x2=rep(.5,40^2),x3=rep(.5,40^2))
## predict, termwise
pv <- predict(b,newdata=pd,type="terms")
colnames(pv)
## plot result
contour(zm,xm,matrix(pv[,1],40,40))
On Wednesday 25 July 2007 11:43, Lucia Zarauz wrote:

  
    

      
      
    

  


  


      

    

    

      
      

        


  

        

      Lucia Zarauz
    

    

      
      #
    

  


  

      
Hi again,

I have found a post in R.help archives made from someone who had the same problem when exporting the smooth function estimate got from a GAM, to plot it into another graphic software (http://finzi.psych.upenn.edu/R/Rhelp02a/archive/23136.html)

However, I haven't found any reply to this post. 

She suggested the function preplot() and I have hound that function in gam package, but not in mgcv package. Is it possible to do this in mgcv package or do I have to move to gam package?

Thank you very much

Luc?a zarauz


_____________________________________________ 
AZTI - TecnaliaLucia Zarauz 
AZTI - Tecnalia / Unidad de Investigaci?n Marina
Herrera kaia portualdea z/g
20110 Pasaia (Gipuzkoa)
Tel: 943 004 800 - Fax: 943 004 801
e-mail: lzarauz at pas.azti.es
www.azti.es ; www.tecnalia.info	 

_____________________________________________ 

  

**************** LEGE OHARRA **************** AVISO LEGAL **************** DISCLAIMER ****************


Mezu hau pertsonala eta isilpekoa da eta baimenik gabeko erabilera debekatua dago legalki. Jasotzailea ez bazara ezabatu mezua, bidali eta kontserbatu gabe. 


Este mensaje es personal y confidencial y su uso no autorizado est? prohibido legalmente. Si usted no es el destinatario, proceda a borrarlo, sin reenviarlo ni conservarlo.


This message is personal and confidential, unauthorised use is legally prohibited. If you are not the intended recipient, delete it without resending or backing it.

-----Mensaje original-----
De: Lucia Zarauz 
Enviado el: martes, 24 de julio de 2007 13:53
Para: 'Henrique Dallazuanna'
CC: r-help at stat.math.ethz.ch
Asunto: RE: [R] plotting gam models

Hi Henrique,

Thank you for your suggestion. 

Actually, I have already tried it, but I am confused because the plot I get is not the same as the output of plot(model) or plot.gam(model). The yaxis is different

On the other hand, if I build a more complex model, as for example:

model<- gam(x ~ s(lat,long) + s(temperature))

I would like to extract the information to build the effects for each explanatory factor (one graph for s(lat,long) and another for s(temperature)), as R does when you use 'plot(model)' and you press return for subsequent pages.

My final aim is to plot the influence of s(lat,long) as a contourplot superposed on a geographical map. Maybe there is an easier way to do it...

Thank you very much

luc?a



_____________________________________________ 
Lucia Zarauz 
AZTI - Tecnalia / Unidad de Investigaci?n Marina
Herrera kaia portualdea z/g
20110 Pasaia (Gipuzkoa)
Tel: 943 004 800 - Fax: 943 004 801
e-mail: lzarauz at pas.azti.es
www.azti.es ; www.tecnalia.info
_____________________________________________ 
? 
**************** LEGE OHARRA **************** AVISO LEGAL **************** DISCLAIMER ****************
Mezu hau pertsonala eta isilpekoa da eta baimenik gabeko erabilera debekatua dago legalki. Jasotzailea ez bazara ezabatu mezua, bidali eta kontserbatu gabe. 
Este mensaje es personal y confidencial y su uso no autorizado est? prohibido legalmente. Si usted no es el destinatario, proceda a borrarlo, sin reenviarlo ni conservarlo.
This message is personal and confidential, unauthorised use is legally prohibited. If you are not the intended recipient, delete it without resending or backing it.