Having consulted a couple of articles on FDR calculations , I was wondering if someone could point me in the right direction concerning plotting an FDR threshold line on a Manhattan plot. My intention is to plot the FDR line on a Manhattan plot generated with un-adjusted p-values. I used python stats-model for FDR adjustment hence, I know the total number of tests performed, I have an estimate of the number of rejected tests (from stats model), and I adjusted for an alpha of say 0.05.
I assume the FDR value will be a scalar quantity based on calculations. Is my assumption right?
Also, once the FDR value is calculated, how is it translated into p-value space to be plotted? for example, is it -log10 (FDR) or is there an FDR line equation to use?
An example of a Manhattan plot with the FDR line can be found here (figure 2) :
http://www.oncotarget.com/index.php?journal=oncotarget&page=article&op=view&path[]=11436&path[]=36221
Find the largest (unadjusted) p-value for which the FDR is below the desired level. Draw the line at that value of p.
PS: Don't confuse alpha and FDR. They coincide only if the "family" of tests has a size of one.
It would be better to plot the adjusted p-values and a line at the desired cutoff. The largest unadjusted p-value that coincides with the FDR cutoff won't be at the exact threshold.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Meta-Analysis