Fisher’s exact test for an r x c table tests for association between two variables. An r x c table is one where the variable assigned to the rows has r levels and the variable assigned to the columns has c levels. Fisher’s exact test assumes that the row and column totals are fixed, and uses the hypergeometric distribution to compute probabilities.
This test does not depend on any large-sample distribution assumptions. This means that it is appropriate for situations where the likelihood ratio and Pearson tests become less reliable, such as small sample sizes or sparse tables.
The Fisher’s Exact Test report includes the following information:
Table Probability (P)
The probability for the observed table. This is not the p-value for the test.
Two-sided Prob ≤ P
The significance probability (p-value) for the two-sided test.
For 2 x 2 tables, the Fisher’s Exact test is automatically performed, unless one row or column contains all zeros (in this case, the test cannot be calculated). See Tests Report.