FisherExactTest dies horribly on corner case inputs

[rohanmaddamsetti]

Hello,

FisherExactTest() dies horribly if you supply matrices like the following: FisherExactTest(5, 0, 5, 0).

Of course one could say "don't do that", which is reasonable when running statistics interactively, but users like myself may want to run a series of statistical tests on data within a loop, without having to write error-checking code for the inputs going into FisherExactTest().

It would be great if this julia function could gracefully report the outcome rather than dying with an ArgumentError. For comparison, R reports the following given the same input.

fisher.test:
data:  TestMatrix
p-value = 1
alternative hypothesis: true odds ratio is not equal to 1
95 percent confidence interval:
   0 Inf
sample estimates:
odds ratio 
         0 

Thanks for your consideration, Rohan

[nalimilan]

Indeed. We should detect parameter values which would throw an error if passed to FisherNoncentralHypergeometric as we do now, and instead fill the FisherExactTest object with values similar to what R does. Any help welcome.

[ararslan]

Fixed by #279