A multiple testing procedure for testing several groups of
hypotheses is implemented. Linear dependency among the hypotheses
within the same group is modeled by using hidden Markov Models.
It is noted that a smaller p value does not necessarily imply
more significance due to the dependency. A typical application is
to analyze genome wide association studies datasets, where SNPs
from the same chromosome are treated as a group and exhibit
strong linear genomic dependency. See Wei Z, Sun W, Wang K,
Hakonarson H (2009)