[BioC] Deciding on a cut off after QC

[Gordon Smyth]

At 02:58 PM 17/05/2005, Ankit Pal wrote:
>Dear Dr Smyth,
>I would prefer using the p-value with a threshold
>value of 0.05.
>In the case of an experiment(I sent you the code) I
>did, the reult file contains p-values (after fdr) in
>range of 0.96 - 0.97.

Clear result. No selected genes. End of story.

>How did I get such values and where do I place a cut
>off in this case?
>Am I doing something wrong?
>In the user guide it says
>"If none of the raw p-value are less than 1/G,
>where G is the number of genes included in the fit,
>then all of the adjusted p-values will the
>equal to 1.Since 1/G is about the expected size of the
>smallest p-values given purely random variation and
>uniform p-values, this means that there
>is no overall evidence of differential expression."
>
>I am not too sure I understand what it means.
>Should there be atleast one p-value < 1/G?

No there need not be. The above text which you have quoted already tells 
you so. What would be the meaning of statistical significance, if you were 
guaranteed to get a significant result in every single experiment that you did?

>I am attaching a sample data output file to this mail.
>What should be done in my case?

Perhaps take a more sophisticated approach to background correction and 
spot filtering. Or else accept that you don't have statistical significance 
for this data.

Gordon

>Thank you,
>sincerely,
>-Ankit