[Rd] Mysterious behavior

Therneau, Terry M., Ph.D. therneau at mayo.edu
Fri Jul 9 16:55:40 CEST 2010


I had trouble with some tests for the survival suite last night that I
cannot explain. 

Framework: Ubuntu Linux, R2.11.  
   For testing survival I have a separate directory and Makefile.  I
pull everything into the local .RData, no packages, library, or
namespace.  (It's easier to add test modifications to a routine in a
chain of calls).  
   
   A test of survreg + psline would fail because splines is not
automatically loaded in this scenario.  In my console window type
	> library(splines)
      > fit <- survreg(Surv(time, status) ~ ph.ecog + spline(age),
data=lung)
as the first 2 lines a a new session (the lung data is also saved)
and up would pop an error message about having <2.10 version of
survival.  And it is true that I had such a thing in my ~/RLib.  But
there is no mention of either the local or the system survival library
when I type search(), and .Autoloaded is NULL. 

  I'm mystified as to how R decided to use that library, and what it was
trying to pull from it.  I removed it from ~/Rlib and the message went
away, but now I'm not so sure that this was a good idea.  Is something
now being automatically snatched out of the standard survival library?
If so that compromises the integrity of my testbed.  How do I find out?

Terry Therneau



More information about the R-devel mailing list