R help - Apr 2009 - Problem with survival

It is likely a problem with survival, since
 	2.9 merged in a large number of changes that had occured in my source 
tree that had not propogated to the R tree
 	
 	my test suite doesn't have a test for this particular case (2 factors)
 	
 	and - Murphy's law applies: although almost every possible case is 
covered in the test suite, any new error will hit an omitted combination of 
options with high probability.
 	
 I won't get to it for a few days though.  As with other errors it will
result
in both a fix and an addition to the test suite.
 
 Thank you for a clear explanation of the problem.  
 
 	Terry T.
 	

Dieter Menne wrote:
 "Make sure that this is really a problem with different versions of R,
not a problem of different versions of survival which was changed recently,
without backward compatibility, so that for example many function of
Design (Harrell) do not work currently."

  The comment about backwards compatability is a little unfair.  The code for 
survival curves post Cox model finally added the (long requested) ability to 
accomodate case weights.  This added an argument to a C routine.  The Design 
package called my C routine directly. I was not aware of this, there is no 
promise in any R package that the not-meant-to-be-called-by-others C routines 
won't change, Frank H was told about this as soon as we found out, and
he's
working on it.

Glad we know where the problem lies now.
Dimitri

On Tue, Apr 28, 2009 at 11:55 AM, Terry Therneau <therneau at mayo.edu>
wrote:
> ?It is likely a problem with survival, since
> ? ? ? ?2.9 merged in a large number of changes that had occured in my
source
> tree that had not propogated to the R tree
>
> ? ? ? ?my test suite doesn't have a test for this particular case (2
factors)
>
> ? ? ? ?and - Murphy's law applies: although almost every possible case
is
> covered in the test suite, any new error will hit an omitted combination of
> options with high probability.
>
> ?I won't get to it for a few days though. ?As with other errors it will
result
> in both a fix and an addition to the test suite.
>
> ?Thank you for a clear explanation of the problem.
>
> ? ? ? ?Terry T.
>
>
> Dieter Menne wrote:
> ?"Make sure that this is really a problem with different versions of
R,
> not a problem of different versions of survival which was changed recently,
> without backward compatibility, so that for example many function of
> Design (Harrell) do not work currently."
>
> ?The comment about backwards compatability is a little unfair. ?The code
for
> survival curves post Cox model finally added the (long requested) ability
to
> accomodate case weights. ?This added an argument to a C routine. ?The
Design
> package called my C routine directly. I was not aware of this, there is no
> promise in any R package that the not-meant-to-be-called-by-others C
routines
> won't change, Frank H was told about this as soon as we found out, and
he's
> working on it.
>
>
>
>


-- 
Dimitri Liakhovitski
MarketTools, Inc.
Dimitri.Liakhovitski at markettools.com

Terry Therneau wrote:
>  It is likely a problem with survival, since
>  	2.9 merged in a large number of changes that had occured in my source 
> tree that had not propogated to the R tree
>  	
>  	my test suite doesn't have a test for this particular case (2
factors)
>  	
>  	and - Murphy's law applies: although almost every possible case is 
> covered in the test suite, any new error will hit an omitted combination of
> options with high probability.
>  	
>  I won't get to it for a few days though.  As with other errors it will
result
> in both a fix and an addition to the test suite.
>  
>  Thank you for a clear explanation of the problem.  
>  
>  	Terry T.
>  	
> 
> Dieter Menne wrote:
>  "Make sure that this is really a problem with different versions of
R,
> not a problem of different versions of survival which was changed recently,
> without backward compatibility, so that for example many function of
> Design (Harrell) do not work currently."
> 
>   The comment about backwards compatability is a little unfair.  The code
for
> survival curves post Cox model finally added the (long requested) ability
to
> accomodate case weights.  This added an argument to a C routine.  The
Design
> package called my C routine directly. I was not aware of this, there is no 
> promise in any R package that the not-meant-to-be-called-by-others C
routines
> won't change, Frank H was told about this as soon as we found out, and
he's
> working on it.
Exactly, and Thomas Dupont is modifying our code to take advantage of 
weights.  Thanks Terry.  -Frank
> 
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
> 

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