[R] logistic regression 3D-plot

[Frank E Harrell Jr]

heikz at gmx.de wrote:
> Dear R-helpers,
> 
> I tried to create a 3D surface showing the interaction between two
> continuous explanatory variables; the response variable is binary (0/1).
> 
> The model is:
> 
> model<-glm(incidence~sun*trees,binomial)
> 
> then I used "wireframe" to create a 3D plot:
> 
> xyz<-expand.grid(sun=seq(30,180,1),trees=seq(0,4000,10))
> 
> xyz$incidence<-as.vector(predict(model,xyz))
> 
> wireframe(incidence~sun*trees,xyz,scales=list(arrows=FALSE))
> 
> which gives me a 3D plot, but the scaling of the y-axis is wrong. the range
> is not from 0 to 1.
> so my question: is there a way to plot these kind of models, with binary
> response variables?
> 
> thanks for your help, Heike
> 
library(Design)
d <- datadist(mydata); options(datadist='d')
f <- lrm(incidence ~ sun*trees)  # lrm is for binary or ordinal response
plot(f, sun=NA, trees=NA)
# add method='image' or 'contour' to get other types of graphs
plot(f, sun=NA, trees=NA, fun='plogis')  # probability scale

-- 
Frank E Harrell Jr   Professor and Chair           School of Medicine
                      Department of Biostatistics   Vanderbilt University