clarification question

[devonorourke]

Thanks for creating and maintaining a great piece of software. As a very novice user of q values and statistics generally I was hoping you could help me understand the difference between the summary and plot functions outputs. In the vignette that uses the hedenfalk dataset, a portion of the summary output looks like this:

##Cumulative number of significant calls:####           
          <1e-04 <0.001 <0.01 <0.025 <0.05 <0.1
p-value       15     76   265    424   605  868
q-value        0      0     1     73   162  319
local FDR      0      0     3     30    85  167

My takeaway from that table was that for a given FDR level (the columns), we would expect N significant sites (rows) for the p and q-values listed. For instance, at and FDR level of 0.05, we might expect 605 significant p-value tests, while just 162 positive q-value tests.

I'm unclear what the last row, the local FDR, is indicating. These values in this vignette are usually about half of the value reported in the q-value of a similar column (I see the same effect in my own data). At first I thought this represented the estimated number of false positives, but then after reading on in the vignette, you show how to find this answer exactly in section 5.4 with the plot function.

However, if I look at that plot in the vignette, I noticed that at the point where there are about 160 significant tests, we'd expect only about 5 false positives, not the ~85 that are indicated in the table in the local FDR row...

Thus it seems like my understanding of what the local FDR row in the summary output is incorrect.

Thanks for any insights you can offer,

Devon

[jdstorey]

https://jdstorey.org/fas/multiple-testing.html#local-fdr http://genomics.princeton.edu/storeylab/papers/Storey_FDR_2011.pdf