How to use the qvalue function for multiple testing?

Question

[halo22]

Hello All,

This might sound like a very dumb question but I would really like to have your input on this one. I have ran a comparison test on my protein analytes, since there are over 1000+ proteins I would like to correct for multiple comparisons. I understand I can always use the FDR approach(p.adjust) but for the current analysis I would like to use qvalues, I also understand that the results that I get from FDR and qvalues would be very different.

My question is particularly about using the qvalues package. I would like to know how most of you use it.

Currently, I am using the default way qobj <- qvalue(p) but I have seen cases were people provide the lambda, fdr.level, pi.0 method etc. it seems that most of these are optional too. I want to know when to use the default way and when to add the remaining parameters. Is it safe to consider that the default way is correct?

Highly appreciate all your inputs on this, please let me know if the question needs more clarity.

Answer 1

[Ming Tommy Tang]

read here for some idea http://crazyhottommy.blogspot.com/2015/03/understanding-p-value-multiple.html also type ?qvalue to see some help.

Answer 2

[theobroma22]

You should also read the references provided in the qvalue package to get a better understanding of the theory, concept and terms like lambda and pi.0.

Answer 3

[Nicolas Rosewick]

First of all you should plot the distribution of nominal p-values. It's always a good idea to check if the test (and H0) you used is ok. Check here for more information about different p-value distribution : http://varianceexplained.org/statistics/interpreting-pvalue-histogram/

[halo22]

Great post it was very helpful. Thank you!