[Rd] negative p-values from fisher's test (PR#7801)

[Kurt Hornik]

>>>>> mnason  writes:

> Full_Name: Martha Nason
> Version: 2.0.1
> OS: Windows XP
> Submission from: (NULL) (137.187.154.154)

> I am running simulations using fisher's test on 2 x c tables and a
> very small p.value from fisher's test (<2.2e-16) is returned as a
> negative number. Code follows.

>> set.seed(0)
>> nreps.outer <-7
>> pvalue.fisher <- rep(NA,nreps.outer)
>> 
>> population1 <- c( rep("A",300),seq(1:100)) 
>> 
>> population2 <- c( rep("A",100),seq(101:200))
>> 
>> 
>> for (j in 1:nreps.outer){
> + n1 <- sample(30:100,1)
> + n2 <- sample(30:100,1)
> + 
> + group1 <- sample(population1, n1, replace=T)
> + group2 <- sample(population2, n2, replace=T)
> + 
> + pvalue.fisher[j] <-
> fisher.test(table(c(group1,group2),c(rep("group1",n1),rep("group2",n2))))$p.value
> + 
> + print(c(j,pvalue.fisher[j]))
> + 
> + }
> [1] 1.000000e+00 3.581362e-05
> [1] 2.0000000 0.1424779
> [1] 3.0000000 0.1196600
> [1] 4.000000000 0.004222897
> [1] 5.000000e+00 3.234016e-07
> [1] 6.000000000 0.003240286
> [1]  7.000000e+00 -3.847298e-05

>> 
>> fisher.test(table(c(group1,group2),c(rep("group1",n1),rep("group2",n2))))

>         Fisher's Exact Test for Count Data

> data:  table(c(group1, group2), c(rep("group1", n1), rep("group2", n2))) 
> p-value < 2.2e-16
> alternative hypothesis: two.sided 

>> fisher.test(table(c(group1,group2),c(rep("group1",n1),rep("group2",n2))))$p.value
> [1] -3.847298e-05

Thanks.

This comes from

L300:
    /* Process node with new PASTP */
    if (pastp <= obs3) {
	/* Update pre */
	*pre += (double) ifreq * exp(pastp + drn);

in f2xact which (e.g. by adding

	REprintf("pre: %f %d %f \n", *pre, ifreq, (double) ifreq * exp(pastp + drn));

shows that ifreq may have negative values.

Now ifreq comes from ifrq which has

     IFRQ   - Vector of length LDSTP containing the past path
	      frequencies.  					(in/out)

so I am not sure whether we should have negative values here (we migth
be indexing something one off) ...

Does anyone know how the algorithm works in detail?

-k