Alternative to Fisher's Exact for larger than 2x2 contingencies with expected value <6 in >20% of cells

#1
Hi there,

Does anyone have any suggested solutions to the above question?

I have a large 2x8 contingency, which unfortunately has 4 cells with an expected value of less than 5, and therefore can't use a Pearson Chi sq test.

I tried to use the Fisher's test with a Freeman-Halton extension in R, but my computer cannot handle the volume of the test (despite increasing the workspace).

Would a Chi square test with a Yates' correction deal with the issue? (I've always been puzzled as to when to use Yates', as I've been suggested that it is obsolete/superfluous.)

Any other tests?

(I tried to do a search on the forum to no avail - hopefully this is not a repeat!)

Thank you :)
 

hlsmith

Less is more. Stay pure. Stay poor.
#2
I read awhile back that the issue of using chi-sq with low cell expectancies was never really well defined and is likely only an artifact. Meaning that the approach can be used more often than most realize.