Stefan - 2010-09-03

Hello,
I run a LIMMA time course analysis (significance level<0.05) on a dataset with 2 conditions and 4 timepoints. I received 9 different lists of significant genes:
1) condition 1, t1
2) condition 1, t2
3) condition 1, t3
4) condition 2, t1
5) condition 2, t2
6) condition 2, t3
7) condition 1, all
8) condition 2, all
9) all, all

As expected, a p-value cut-off of 0.05 was set in lists 1, 2, and 3 for the respective pair-wise comparisons (t1 vs t2 for list1, t1 vs t3 for list 2, t1 vs t4 for list 3).
I expected the same for lists 4, 5, and 6 (condition 2) but, although there seems to be some kind of filtering, p-values range from very low values to 1. The same is true for the lists 7, 8, and 9.

Is there a dependency  of condition 1 and 2 for the calculation of the p-values for these subdatasets, or is this a bug?

Thanks for you help in advance

Stefan