R help - Jul 2005 - exact conditional mantelhaen.test estimate is 0 ?!

Dear listers,

I am trying to compute the exact conditional test given strata margins
of a 2 by 2 by K array using the mantelhaen.test function to get a
common odds ratio estimate.
The estimate for the test on the following data is 0, which in my
opinion dosen't make any sense.

x <- array(c(53, 6098, 1006, 4521, 63, 8070, 1163, 6137), dim=c(2,2,2))
 x
, , 1

     [,1] [,2]
[1,]   53 1006
[2,] 6098 4521

, , 2

     [,1] [,2]
[1,]   63 1163
[2,] 8070 6137

mantelhaen.test(x, exact=TRUE)
      Exact conditional test of independence in 2 x 2 x k tables

data:  x
S = 116, p-value < 2.2e-16
alternative hypothesis: true common odds ratio is not equal to 1
95 percent confidence interval:
 0.0000000000 0.0001190186
sample estimates:
common odds ratio
                0

Could this be some kind of memory overflow issue since some of the
values are quite high? Or am I completely off here? 

I would appreciate any help on this,

Florian 

-- 
Florian Hahne <f.hahne at dkfz-heidelberg.de>

On Fri, 8 Jul 2005, Florian Hahne wrote:
> Dear listers,
>
> I am trying to compute the exact conditional test given strata margins
> of a 2 by 2 by K array using the mantelhaen.test function to get a
> common odds ratio estimate.
> The estimate for the test on the following data is 0, which in my
> opinion dosen't make any sense.
>
Indeed it doesn't.  Some probabilities are underflowing to zero.  It is 
informative to look at the estimates from
    mantelhaen.test(round(x/s), exact=TRUE)
for various values of s.  The estimate is stable at about 0.04 for 
large s until s gets to just above 1.8, then quickly decreases to zero.

I think you need exact=FALSE for data sets this large.

 	-thomas