[Rd] R (development) changes in arith, logic relop with 0-extent arrays

Therneau, Terry M., Ph.D. therneau at mayo.edu
Wed Sep 7 14:22:01 CEST 2016



On 09/07/2016 05:00 AM, r-devel-request at r-project.org wrote:
> I've been slightly surprised (or even "frustrated") by the empty
> reaction on our R-devel list to this post.
>
> I would have expected some critique, may be even some praise,
> ... in any case some sign people are "thinking along" (as we say
> in German).

Have patience Martin.  I read the news in digest form once a day and just saw the 
announcement and this follow-up a few minutes ago.  I haven't had cause to download a new 
copy of R-devel in the last month so the new Sept 5 version hasn't impacted me.

The survival package had one line of code that will need modification, which isn't so bad. 
  It involves the scaled Shoenfeld residuals, for which the formula is r V + 1 beta'; r= 
matrix of residuals with nvar columns and #deaths rows, V=vcov(fit) = variance matrix, and 
then add beta[j] to each column j.  When the number of covariates is 1 the first term of 
this collapsed to a vector *  1x1 matrix, mostly because of how the code was written, 
which was itself the result of evolution over time.   Bottom line is that I don't find 
either the prior behavior or the new one problematic.

Terry T.



More information about the R-devel mailing list