Question: How To Verify That The False Discovery Rate Qvalues Are Correct?
gravatar for tetardjones
7.6 years ago by
Newcastle University
tetardjones0 wrote:

I've just started using R and fdrtool, and I'm not sure if the qvalues I'm receiving back are accurate. I performed fdrtool on pvalues obtained from a two way anova on proteomics data. So I have 266 data values (protein spots) for two factors (ft, vr, and the interaction) for each biological sample.

One of the two factors (vr) has a highly significant effect with 119 protein spots significantly affected at p<0.05. When I run the fdrtool, the qvalues are slightly higher than pvalues (as expected), but only up to p<0.01. Between p<0.01 and p<0.05, the qvalues are lower, giving me more significant protein spots at that level - is this correct?

The other factor (ft) had only 3 weakly significant protein spots. When I run fdrtool, all 266 qvalues are 1.

The interaction effect (ftxvr) produced 14 significant pvalues (mostly p<0.05, a couple are p<0.01). fdrtool produces qvalues ranging between 0.87-0.99 and the rise with rising pvalues, so I lose the significant results here.

Best wishes, Catherine

R • 2.0k views
ADD COMMENTlink modified 7.6 years ago by Istvan Albert ♦♦ 85k • written 7.6 years ago by tetardjones0

Mmmm, I am not sure, but I would expect that q-values should be >= p-values, always. Maybe you can post some small example dataset giving "strange" results. Also, you could try using p.adjust. It always worked well for me.

ADD REPLYlink written 7.4 years ago by Fabio Marroni2.6k
Please log in to add an answer.


Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 2.3.0
Traffic: 1738 users visited in the last hour