$begingroup$
I have dataset of 2 time points. In each time point I have to compare the counts of different components at each position (number of positions =1000). I used fisher exact test to compare if there is a significant difference in the counts of the different components at the 2 time points. I ended up performing 1000 fisher test then I applied FDR correction. My questions are:
- If the total number of samples used to compare between the 2 time points was varying significantly as in the picture below how that would affect the results? and the p value ? are there any bias?
- How to fix this to get reliable result or is there any way for normalization to account for this variation when performing fisher exact test ?
This is an example for more clarification, so for each position (each row in this image) I performed fisher exact test and then applied FDR correction:
Any help is appreciated.
$endgroup$
0