How to deal with zero cells in 2 x 2 tables, using R I am trying to look at the relationship between two nominal variables, using R. However, I am not certain how to deal with cases where a 2 x 2 contingency table has one or more zero cells, as shown below:


The two variables are the number of N2 and that of the verb. For instance, "A lot of money is being spent from certain countries" is a case that fits the cell of "N2nonpl" and "Vsing" in the first contingency table. I tried Fisher Exact Test with the very simple following codes:
testor1 = rbind(c(170,0),c(1,1050))
fisher.test(testor1)

testor2 = rbind(c(0,0),c(49,34))
fisher.test(testor2)

Are there any easy, proper ways of dealing with such examples, using R?
Thank you very much in advance!
 A: A single zero cell is not of itself of major consequence; nor even two in a 2x2 table if they don't share a row or column. That is, case 1 should not present any issue - your estimate of the odds ratio should be infinite and your p-value should be almost 0.
However two 0's in the same row or column is problematic. If you condition on the margins (as is the case with the Fisher test), there's only one possible table consistent with both margins; your p-values will always be 1. In those specific circumstances there's nothing to be done about that, there's no information about relative proportions in the zero row (or column, as appropriate).
Unconditionally you might just about be able to have some way to get to a p-value that's not exactly 1 (I haven't tried to come up with a way that would) -- but even if you could, it's really not going to help you much; the uncertainty in the proportions will swamp the effect in the mean (i.e. I believe you would still not be able to tell the relative proportions apart, since the uncertainty around any estimates would be almost entirely overlapping).
