I am curious. Having done big data approaches and smaller experiments in biology, I am wondering whether there is a good reason for us to use FWER corrections of datasets with fewer test numbers (e.g. less than 10) rather than FDR which we use typically in large datasets. In a smaller number of tests using an FDR, we are still controlling the level of false positive discovery but significantly increase the error for false negatives. So why is it that most biology related papers (I am working in the field of immunology) are using Bonferroni or Tukey or Sidak rather than FDR? Is it just a question of stringency? What about the number of discoveries we loose because of the reduction of power in FWER corrections?
Felix Clemens Richter it is safe to assume that all tests are exchangeable the power to detect discoveries is equally likely among all tests.
FDR methods increase power by incorporating informative covariates. However, there is currently no consensus on how the modern methods compare to one another. The answer to which methods can be used for a particular analysis is often unclear. FWER are highly conservative, controlling (type I errors) greatly reduced power. The FDR was more recently proposed as an alternative metric to the FWER in multiple testing control.
Good luck
I think that this attached file will be helpful to you. Best wishes, David Booth
Oops I almost forgot. See Steen Knudsen, Cancer Diagnostics with DNA Microarrays, available in the z-library. This answers the question about missing things using multiple comparison methods. Best wishes, David Booth
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics