exact test for large tables. Was: [R] unexpected R crash - again

[Yudi Pawitan]

If you have an actual large table to analyse, rather 
than trying to solve the space problem, you may want 
to consider a MonteCarlo implementation 
of the exact test. Very easy to implement in R. See, 
for example, Lange's Numerical Analysis for Statisticians, 
Section 21.7.

-Yudi-


At 04:10 PM 8/25/00 -0500, Andy Jaworski <apjaworski at mmm.com>
 wrote:
>Sorry, but I lost this thread, so I sending this as a new message.
>
>This is really a follow-up to a post from a couple days ago saying that
>fisher.test from the ctest library crashed on the following data set:
>> T
>      [,1] [,2]
> [1,]    2    1
> [2,]    2    1
> [3,]    4    0
> [4,]    8    0
> [5,]    6    0
> [6,]    0    0
> [7,]    1    0
> [8,]    1    1
> [9,]    7    1
>[10,]    8    2
>[11,]    1    0
>[12,]    3    1
>[13,]    1    1
>[14,]    3    0
>[15,]    7    2
>[16,]    4    1
>[17,]    2    0
>[18,]    2    0
>[19,]    2    0
>
>Then Professor Ripley responded that the problem was not related to the
>Windows port (as the original author implied) but to the amount of memory R
>was started with.
>
>I just ran some tests on this I have to report that I am getting crashes
>every time I try this problem.
>(1) On my WinNT machine I got up to
>         Rgui.exe --vsize=220M --nsize=4000k
>     which produces
>> gc()
>           free    total  (Mb)
>Ncells  3857498  4000000  76.3
>Vcells 28790374 28835840 220.0
>
>and it still crashes with the "referenced memory cannot be read" message.  I
>also noticed that the error message comes immediately, no matter how large
>the
>memory allocation is.  This machine is a Pentium Pro 200Mhz with 192Mb of
>memory
>
>(2) On my Linux box I can only go to about --vsize=120M --nsize=1000k.  This
>is a Pentium MMX 266Mhz with 64Mb of memory.  The machine crashes
>     after "thinking" for about 3 seconds.  It produces segmentation
>violation error and dumps core.
>
>It seems to me that, unless I have some unlucky machines, the problem is not
>just handling too small memory allocation ungracefully.
>
>Any comments will be appreciated,
>
>Andy
>
>
>-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
.-.-
>r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
>Send "info", "help", or "[un]subscribe"
>(in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch
>_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._
._._
>
Yudi Pawitan     yudi at stat.ucc.ie
Department of Statistics UCC
Cork, Ireland
Ph   353-21-490 2906
Fax 353-21-427 1040 
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._