22 Oct
2010
22 Oct
'10
5:53 p.m.
Hi, I have a dataset that seems to have some very nice group differences (please see the graph and qdec image), but they never seem to survive FDR. I noticed when I run the "Set Using FDR", it generates a very high threshold above 5; where as, with some of my other datasets FDR is set much lower (often in the 2's and 3's). Can you please give me some insight into why the threshold is being set so high? Is it solely linked to the small sample size, or is there something additional we can do in the processing, so that regions with such large group differences will survive correction?
Thank you, Mira