On Thursday, July 20, 2017 7:41:00 PM EDT Duncan Murdoch wrote:
> Thanks for posting this series of patches. Unfortunately, there's a
> good chance they'll get lost in all the traffic on R-devel. If you
> don't hear that they've been fixed in the next couple of weeks, could
> you post them to bug
Thanks for posting this series of patches. Unfortunately, there's a
good chance they'll get lost in all the traffic on R-devel. If you
don't hear that they've been fixed in the next couple of weeks, could
you post them to bugs.r-project.org, and post future patches there as well?
In examples
> On 20 Jul 2017, at 19:46 , Scott Kostyshak wrote:
>
> On Thu, Jul 20, 2017 at 04:21:04PM +0200, Martin Maechler wrote:
>>> Scott Kostyshak
>>>on Thu, 20 Jul 2017 03:28:37 -0400 writes:
>>
From ?confint:
>>> "Computes confidence intervals" and "The default method assumes
>>>
Hello,
This is a patch to fix what appears to be a simple typo. The warning says
"invalid status assuming 0", but then instead sets runLast to 0.
Signed-of-by: Steve Grubb
Index: src/main/main.c
===
--- src/main/main.c (revisio
Hello,
The SMBUF_SIZED_STRING allows fscanf to read upto 511 bytes. The buffer
at line 1382 is only 128 bytes. The fscanf format specifier ought to be
resized to prevent a stack overrun.
Signed-of-by: Steve Grubb
Index: saveload.c
Hello,
This patch fixes a memory leak due to ptd going out of scope
before its assigned to dd.
Signed-off-by: Steve Grubb
Index: src/library/grDevices/src/devPicTeX.c
===
--- src/library/grDevices/src/devPicTeX.c (revision 72
Dear Joris,
I’ll be more careful in my wording next time; thanks for the pointer, and
thanks for the discussion. This whole process has been quite educational! 😉. I
think we’ve reached a consensus here, where the situation as it is right now
has been chosen to allow for flexibility of R’s glm()
Hello,
There are times when b points to buf which is a stack variable. This
leads to a bad free. The current test actually guarantees the stack
will try to get freed. Simplest to just drop the variable and directly
test if b should get freed.
Signed-off-by: Steve Grubb
Index: src/main/connec
Hello,
There appears to be a break missing in the switch/case for the LISTSXP case.
If this is supposed to fall through, I'd suggest a comment so that others
know its by design.
Signed-off-by: Steve Grubb
Index: src/main/builtin.c
On Thu, Jul 20, 2017 at 8:32 PM, Harm-Jan Westra wrote:
> My apologies if I seemed to ‘blame R’. This was in no way my intention. I
> get the feeling that you’re missing my point as well.
>
I get that now. But you're on R-devel and you started with the claim that R
"falsely reports...". That loo
Dear Mark,
I agree that convergence is a problem that applies to optimization in general,
where the function you�re trying to optimize may have more than one local
minimum. In your case, you probably would have to try different starting points
for the MLE procedure. This should not be the case
Hi Harm-Jan. I've been following this thread to some degree and just want
to add that
this issue is not specific to the GLM. It's a problem with optimization of
functions in general. I was using use Rvmmin with constraints which is an
extremely solid optimization package written by John Nash ( use
My apologies if I seemed to �blame R�. This was in no way my intention. I get
the feeling that you�re missing my point as well.
I observed something that I thought was confusing, when comparing two more or
less identical methods (when validating the C code), and wanted to make a
suggestion as t
Dear Simon,
Thanks for your response. I have a suggestion that could be non-intrusive, but
still provide some additional info to the user.
The glm function already checks for collinearity of the input, and you can
easily check which covariate was aliased as a result, using
summary(model)$ali
On Thu, Jul 20, 2017 at 04:21:04PM +0200, Martin Maechler wrote:
> > Scott Kostyshak
> > on Thu, 20 Jul 2017 03:28:37 -0400 writes:
>
> >> From ?confint:
> > "Computes confidence intervals" and "The default method assumes
> > asymptotic normality"
>
> > For me, a "con
On Thu, Jul 20, 2017 at 6:21 PM, Harm-Jan Westra wrote:
> Dear Joris,
>
>
> I agree that such a covariate should not be used in the analysis, and
> fully agree with your assessment. However, your response assumes that
> everybody who uses R knows what they're doing, which is a dangerous
> assumpt
How should R deal with matrices that have a 'names' attribute? S (and S+)
did not allow an object to have both dims and names but R does. However,
some R functions copy the dims but not the names to the returned value and
some copy both. I don't see a pattern to it. Is there a general rule for
In defence of Harma-Jan's original post I would say that there is a difference
between true convergence and satisfying a convergence criterion.
In my view the algorithm has not converged. This is a case of quasi-complete
separate -- there are both successes and failures when x13=0 but only fail
Dear Joris,
I agree that such a covariate should not be used in the analysis, and fully
agree with your assessment. However, your response assumes that everybody who
uses R knows what they're doing, which is a dangerous assumption to make. I bet
there are a lot of people who blindly trust the
Allow me to chime in. That's an interesting case you present, but as far as
I'm concerned the algorithm did converge. The estimate of -9.25 has an
estimated standard error of 72.4, meaning that frequentists would claim the
true value would lie anywhere between appx. -151 and 132 (CI) and hence the
> nicola farina
> on Thu, 20 Jul 2017 16:51:54 +0200 writes:
> Dear all,
> I just installed R 3.4.1 and noticed a change in how user-defined
functions
> are printed. A small example:
> string<-"f<-function(x){\n\tx^2\n}"
> cat(string,file="tmp00a.R")
> source
I have something working using configure.win, but this changes the
executable when the windows version of the package is created via R CMD
INSTALL --build.
Could there be any way to do so at installation time from the binary
package?
Thanks,
Iago
On 20 July 2017 at 16:21, Iago Mosqueira wrote
Dear R-core,
I have found an edge-case where the glm function falsely concludes that the
model has converged. The issue is the following: my data contains a number of
covariates, one of these covariates has a very small variance. For most of the
rows of this covariate, the value is 0, except f
Dear all,
I just installed R 3.4.1 and noticed a change in how user-defined functions
are printed. A small example:
string<-"f<-function(x){\n\tx^2\n}"
cat(string,file="tmp00a.R")
source("tmp00a.R")
f
And this is what I see:
#R 3.4.0
function(x){
x^2
}
#R 3.4.1
function(x){
\tx^2
}
Seems
Thanks. I did not explain clearly that the executables are not compiled
during package compilation, apologies. They are compiled beforehand, as
they use ADMB (AD Model Builder) and placed in inst/bin/windows.
I assume Makefile.win could copy the appropriate one from PKG/bin/$arch to
PKG/inst/bin/
> Scott Kostyshak
> on Thu, 20 Jul 2017 03:28:37 -0400 writes:
>> From ?confint:
> "Computes confidence intervals" and "The default method assumes
> asymptotic normality"
> For me, a "confidence interval" implies an exact confidence interval in
> formal statistics
On Thu, Jul 20, 2017 at 1:55 PM, Iago Mosqueira
wrote:
>
> I maintain a repository of R packages, where some of them contain
> executable binaries. I need to separate those compiled for 32 and 64 bit in
> Windows.
Have a look at the antiword package. It has a simple makevars which
builds antiword
On 20.07.2017 13:55, Iago Mosqueira wrote:
Hi,
I maintain a repository of R packages, where some of them contain
executable binaries. I need to separate those compiled for 32 and 64 bit in
Windows, but I could not how to do any of the two options I can think of:
1. Have subarch subfolders in
Hi,
I maintain a repository of R packages, where some of them contain
executable binaries. I need to separate those compiled for 32 and 64 bit in
Windows, but I could not how to do any of the two options I can think of:
1. Have subarch subfolders in PKG/inst/bin to that the right one is
installed
Seems like I just need to put /usr/xpg4/bin first in the PATH, that
has a feature full grep program, and then the configure test passes.
Gabor
On Thu, Jul 20, 2017 at 11:19 AM, Gábor Csárdi wrote:
> This is R-patched, from 15th of July. I get:
>
> configure:5117: found /opt/csw/bin/texi2any
> co
This is R-patched, from 15th of July. I get:
configure:5117: found /opt/csw/bin/texi2any
configure:5129: result: /opt/csw/bin/texi2any
configure:5141: checking whether texi2any is at least 5.1
configure:5163: result:no
However:
/opt/csw/bin/texi2any --version
texi2any (GNU texinfo) 6.1
Copyrigh
>From ?confint:
"Computes confidence intervals" and "The default method assumes
asymptotic normality"
For me, a "confidence interval" implies an exact confidence interval in
formal statistics (I concede that when speaking, the term is often used
more loosely). And of course, even if a test statis
32 matches
Mail list logo
