
Compute effective lengths
    http://robpatro.com/blog/?p=235
	Also discusses TPM

Mann-Whitney for odd numbers of replicates coming out slightly higher
than wilcox.test()

Docs: Never make a decision based solely on the p-value and/or fold-change

Report stats on counts, including mean and maximum deviation as % of mean
to help guide pval-sim parameters

Proper boostrap analysis to show variance in near-exact P-values used
for sample counts 5 to 12.  Low priority, since most studies have
either 3 samples, or a high number so that Mann-Whitney is used instead.
