[R] Graphical presentation of logistic regression

[Barry Rowlingson]

Frank E Harrell Jr wrote:

>>I
>>really nead to know how to invert a histogram and how to overlay this
>>with another histogram "the right way up".
>>

  Flipping the last two numbers in par()$usr has this effect...

>  From what you describe, that is a poor way to represent the model 
> except for judging discrimination ability (if the model is calibrated 
> well).  Effect plots, odds ratio charts, and nomograms are better.  See 
> the Design package for details.

 From the poor-diagnostics-R-us dept, here's something to work from:


logDiag <- function(x,y){

   d0=x[y==0]
   d1=x[y==1]

   h0=hist(d0,plot=FALSE)
   h1=hist(d1,plot=FALSE, breaks=h0$breaks)

# set the xlim so the stalactites dont hit the stalagmites:
   plot(h0, ylim=c(0,max(c(h0$counts,h1$counts))*2))
   pu=par()$usr

# flip the Y-axis limits
   par(usr=pu[c(1,2,4,3)])

# draw the stalactites
   lines(h1)

# reset axis
   par(usr=pu)

}

sample:

  xyd=data.frame(x=runif(1000),y=as.numeric(runif(1000)>.2))
  logDiag(xyd$x,xyd$y)

  lets just hope this plot isn't patented like that baseball diamond 
plot a few years ago. My lawyer is ready...

Baz