PermuSeed KS values : pValue / qValue?

[dS]

Hi EGAN,

I ran a lot of PermuSeed enrichments on RNAseq Data comparing 2 conditions. I was wondering, the PermuSeed KS value indicate a p-value I assume. What is the Q-Value?

Thanks

[ucsf egan]

q-values are an estimate of the False Discovery Rate (FDR), given multiple hypothesis tests, assuming no correlation between tests.  As an alternative to Bonferroni and other p-values adjustment methods, they are often reported in journals instead of p-values.

The typical interpretation of q-values/FDR is that if you were to cut off significance at a certain p-value, the q-value would tell you the likely proportion of false positives in the set of significant genes/gene sets.

Modern hypothesis testing is not built for multiple tests, but statisticians are bending over backwards to try to account for them.  Using EGAN, or any software that produces p-values, it is important to remember that the p-value significance is only valid upon corroborating evidence.

Hope that helps,

Jesse

--
You received this message because you are subscribed to the Google Groups "UCSF EGAN" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ucsf-egan+...@googlegroups.com.
To post to this group, send email to ucsf...@googlegroups.com.
Visit this group at http://groups.google.com/group/ucsf-egan.
For more options, visit https://groups.google.com/d/optout.

[dS]

Thanks Jesse,

so would you show the p-value and the q-value (FDR) in a table?

[ucsf egan]

Showing both in the table is a good idea.  I suggest using p-value for coming up with new hypotheses that you might test.

Jesse

[dS]

Thanks Jesse,

I actually decided to plot p vs. q value in a scatter  instead of just using a table.

One more question, I read in another post, that the paraSeed is already a -log10, so does the permuSeed also represent the -log10(p-value) or did I read this wrong?

Best,

Dirk

[ucsf egan]

Hi Dirk,

PermuSEED values range between 0 and one, and are not log10.

Jesse