I'll be very thankful for some feedback here.
I've got a binary matrix of samples X traits, so for each sample I know whether a trait exists or not.
Each sample belongs to a group of samples and I want to test which, if any, of the traits appears dis-proportionally in a group. Some numbers to put things in context- 500 samples, 10 groups, 2000 traits.
I thought about two ways to approach it and will be glad to get your input:
1) Propotion test- test the observed proportion of the trait in a group relative to the expected proportion (the proportion of the samples in the group relative to all samples).
2) fisher's exact test with mx2 contingency table- where m is the number of groups and 2 is four counting samples with or without the trait. Something like that:
My problem here is that in contrast to 2x2 table the p value is not so intuitive to me.
Thanks in advance!