I'm using the timecourse library to calculate Hotelling's T2 statistics, and would like to convert these to P values (and, ultimately, perform FDR). For a 4 time point, 3 replicates (control and treated each), 54K gene experiment, I'm struggling with the appropriate variables for transformation. This paper: http://bioinformatics.oxfordjournals.org/cgi/content/full/21/14/3105 provides an equation (see below) in J (# of genes), and number of replicates in treated (nD) and control (nN), but (incorrectly) leads to negative F values. Thanks to anyone else who has already solved this, -Aaron P.S. Relevant extract from the above paper: (apologies if this doesn't come through well):