I am working with diet data of chicks from a single colony from multiple seasons. The data consists of actual counts and frequency of occurrence (%) of 4 prey groups from 8 seasons extracted from diet samples. I would like to assess if there is a statistical difference between the proportion of prey groups consumed across the seasons. As the expected value of multiple cells is < 5, I opt for a Fisher's exact test, as opposed to a Chi-squared test, despite my contingency table being 4x8, not 2x2.
The contingency table (df) has Season as my X, and Prey groups (Fish, Cephalopds, Crustacean, Carrion) as my Y nominal variables. I use the following code:
fisher.test(df, simult.p.value = TRUE)
If my cell values of the contingency table are FO%, then I get the following warning message:
Warning message: In fisher.test(df4, simulate.p.value = TRUE) : 'x' has been rounded to integer: Mean relative difference: 0.005178792
This message does not appear if I use the count data instead.
MY QUESTION: From searching the literature, I was under the impression that it is possible to use FO%. Is this not the case? Must I rather use count data?
Any assistance will be greatly appreciated.
Many thanks, AK