[R] How to maintain class signature in splom

[Gosink, John]

All,

I was having trouble trying to create a new class of data and pass it on to splom (in the lattice library).   I mentioned this to Martin Morgan after a talk he gave.   Following is not so much a question, but rather an answer from Morgan that might be useful to others.   Here is the edited part of an email conversation with him:


On Thursday, August 26, 2010 1:36 PM, Martin Morgan wrote:

Fun! I think the problem is

> x = new("censored", lung$time, event=lung$status)
> x[1:5]
[1]  306  455 1010  210  883

which splom does (subsetting) as part of it's parsing of the general case. So

setMethod("[", "censored",
          function(x, i, j, ..., drop=TRUE) 
{
    initialize(x, x at .Data[i], event=x at event[i])
})

and

> x[1:5]
An object of class "censored"
[1]  306  455 1010  210  883
Slot "event":
[1] 2 2 1 2 2

which I think works. If so, and if your post is still dangling on R-help, then maybe you could follow up there to help others in the same position?

Martin

 -------------


On 8/26/2010 10:13 AM, Gosink, John wrote: 
Hi Martin,
 
I enjoyed the presentation last night.  As per the brief discussion afterwards, I'm having some trouble getting a new class I created to work its way into splom.   The basic idea is that I want to trick out the upper.panel and lower.panel functions in splom to recognize different flavors (classes) of 'x' and 'y' values and do different behaviors accordingly.   Below is what I posted to the BioConductor mailing list but didn't get any answer.    Perhaps something is obvious to you?   Any advice you could give would be greatly appreciated.


# This is a new class to combine the time and censoring information
# into one column in the data frame that I'll use
setClass("censored",
         representation(event = "numeric"),
         contains = "numeric")
 
 
# This is a simplification of what I'm doing to show that I'm not retaining
# the 'censored' class into the lower.panel function
panel.lowerTest <- function(x, y, ...) {
  print(class(x))
  print(x)
  print(class(y))
  print(y)
  cat("\n\n ---------\n\n")
  panel.xyplot(x,y,...)
}
 
# Here is a simple data example
library(survival)
library(lattice)
data(lung)
lung = lung[1:50,c("inst", "time", "status", "age", "sex")]
lung$sex = as.factor(c("M", "F")[lung$sex])
 
# Create a 'censored' object and tack it onto my data set
lung$OS  = new("censored", lung$time, event=lung$status)
print(lung$OS)  # This is what a 'censored' object should look like
 
# Do the splom.  Notice that the lower panel should have reported that 
# it was seeing 'censored' data, but it didn't...
splom(lung[, c("sex", "age", "OS")], lower.panel = panel.lowerTest)



________________
John Gosink, PhD
Senior Scientist, Amgen, Inc.
1201 Amgen Court West
Seattle, WA 98119-3105
(w) 206 265-8217