R help - Feb 2013 - Cramer von Mises test for a discrete distribution

Hi,
?
I'm trying to carry out Cramer von Mises tests between pairs of vectors
belonging to a discrete distribution (concretely frequencies from 0 to 200).
However, the program crashes in the attempt. The problem seems to be that these
vectors only have positive integer numbers (+ zero). When I add a random very
small positive decimal to the non-decimal part everything works fine (files prm1
& prpmr1). I attach two of these vectors so you can run the following code.
I've also thought to divide both vectors by a real constant such as pi. Do
you think these two approaches are acceptable?
?
setwd("")
require(CvM2SL2Test)
prm = scan('prm.txt')
prpmr = scan('prpmr.txt')
ct1 = cvmts.test(prm, prpmr) # here R crashes
ct1
cvmts.pval( ct1, length(prm), length(prpmr) )
?
?
Thank you for your help,
?
Santi
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: prm.txt
URL:
<https://stat.ethz.ch/pipermail/r-help/attachments/20130219/a543b86c/attachment.txt>
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: prpmr.txt
URL:
<https://stat.ethz.ch/pipermail/r-help/attachments/20130219/a543b86c/attachment-0001.txt>
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: prm1.txt
URL:
<https://stat.ethz.ch/pipermail/r-help/attachments/20130219/a543b86c/attachment-0002.txt>
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: prpmr1.txt
URL:
<https://stat.ethz.ch/pipermail/r-help/attachments/20130219/a543b86c/attachment-0003.txt>

On Tue, Feb 19, 2013 at 2:49 PM, Santiago Guallar <sguallar at yahoo.com>
wrote:
> Hi,
>
> I'm trying to carry out Cramer von Mises tests between pairs of vectors
belonging to a discrete distribution (concretely frequencies from 0 to 200).
However, the program crashes in the attempt. The problem seems to be that these
vectors only have positive integer numbers (+ zero). When I add a random very
small positive decimal to the non-decimal part everything works fine (files prm1
& prpmr1). I attach two of these vectors so you can run the following code.
I've also thought to divide both vectors by a real constant such as pi. Do
you think these two approaches are acceptable?
 >
 > setwd("")
 > require(CvM2SL2Test)
 > prm = scan('prm.txt')
 > prpmr = scan('prpmr.txt')
 > ct1 = cvmts.test(prm, prpmr) # here R crashes

 For you maybe. For me, works fine, and:

 > ct1

[1] 30.20509

 > cvmts.pval( ct1, length(prm), length(prpmr) )

 - this is taking a bit longer. I gave up and killed it. Maybe it
would have eventually "crashed R", but you said the other function
call crashed R.

Your two mistakes are:

 1. Saying "R crashes" without showing us any kind of crash report or
error message.
 2. Not listing your system and package versions.

Ah, your three mistakes are...

 3. Not reading http://www.r-project.org/posting-guide.html


Barry

On Wed, Feb 20, 2013 at 10:03 AM, Santiago Guallar <sguallar at yahoo.com>
wrote:
> Thanks Barry,
>
> Following your list order
> 1) It pops up a window saying R for windows GUI front-end crashed. Below
> three options: look for on-line solutions; shut down the program; debug the
> program (I'm translating from Spanish)
 That's good - often people say "crash" when all they have seen is
a
"stop" from R.
> 2) The processor of my laptop is an Intel Core duo 1,60GHz with ram= 4 GB,
> 32 bits. The R version I have installed is 2.15.2 (2012-10-26)
 Nicely up to date...
> 3) I read the posting-guide. Ok, it may be basic statistics. Question
> withdrawn
 Oh don't do that! You're not asking how to do basic statistics, you
are trying to do it yourself and getting a crash. Fair question for
starters...
> I made an additional mistake: I attached the wrong files. Please run the
> code with these, and you'll see the problem.
 Will I, will I, will I....
> ct1 = cvmts.test(prm, prpmr) # here R crashes
 *** caught segfault ***
address 0x5620e458, cause 'memory not mapped'

Traceback:
 1: .C("CvMTestStat", as.double(x), as.integer(length(x)),
as.double(y),     as.integer(length(y)), testscore = double(1))
 2: cvmts.test(prm, prpmr)

Possible actions:
1: abort (with core dump, if enabled)
2: normal R exit
3: exit R without saving workspace
4: exit R saving workspace
Selection:

Yes! This looks like a bug in that package function, a bit of
investigation seems to blame it on when you have repeated values in
the vectors:
> cvmts.test(1:10,1:10)
[1] 0.025
> cvmts.test(rep(1,10),rep(1,10))
[1] 0.955
> cvmts.test(rep(1,10),rep(2,10))
 *** caught segfault ***
address 0x514daba8, cause 'memory not mapped'

Traceback:
 1: .C("CvMTestStat", as.double(x), as.integer(length(x)),
as.double(y),     as.integer(length(y)), testscore = double(1))
 2: cvmts.test(rep(1, 10), rep(2, 10))

Possible actions:
1: abort (with core dump, if enabled)
2: normal R exit
3: exit R without saving workspace
4: exit R saving workspace

Functions shouldn't crash like this - so time for you to email the
maintainer:
> packageDescription("CvM2SL2Test")$Maintainer
[1] "Yuanhui Xiao <yxiao at gsu.edu>"

The function disappears into C code, but I suspect its dividing by
zero somewhere...

Barry