Barnards test 2x4 table in R I have a table of 2x4 data which I conducted to test for no association. I can do this test fine in R and I get a p value I can interpret.
However I did a bit of googling and I came across another more robust test called Barnard's test. However it seems the current R implementation of Barnard's test library only allows for testing  a 2x2 contingency table.
I'm wondering if it's possible to do a series of Barnard's tests for association between 2 of the columns in turn and have a set of aggregate p values I can interpret? 
Or should I just stick with the Fisher exact test for my 2x4 table?
 A: Yes, just stick with Fisher's exact test. It is the standard statistical test for this problem.
First, Barnard's test is difficult to implement for anything larger than a 2x2 tables, whereas Fisher's test can be evaluated regardless of the number of rows and columns.
Second, Barnard's test is not more "robust" than Fisher's test. Barnard's test makes more assumptions than does Fisher's test (it assumes random margins whereas Fisher's test is valid even if the marginal totals were pre-determined) and so is, if anything, less robust. The only possible advantage of Barnard's test, when it is applicable, is that it can be slightly more powerful than Fisher's test. However the gain in power isn't dramatic, as you can readily verify yourself by trying some examples. This may explain why it is rarefy used. Even Barnard himself recommended against it.
I am not clear why you would want to conduct tests for a series of 2x2 tables. If you did that, then adjusted for multiple testing, you would end up with less statistical power than if you just used Fisher's test for the whole table.
