[BioC] get over it

Chad Shaw cashaw at bcm.tmc.edu
Wed Dec 17 18:33:31 MET 2003


I would like to comment on all this marginal testing/ multiplicity 
adjustment/  Fold change/T test stuff..

My comment is: "let's keep our eyes on the prize": the scientific 
questions about biological systems
that can be addressed with array data.  In my opinion these REAL and 
RELEVANT questions
are NOT so essentially tied to the marginal testing issues for single genes.

In my opinion, the real questions about microarray data ARE NOT single 
gene questions.

Focusing on marginal tests of single genes IS KIND OF A WASTE OF TIME. 
 It the kind of thing
that makes statisticians irrelevant to scientific inquiry.  The kind of 
thing that makes us
accessories rather than scientists.  

IT IS A MISUSE of  MULTIVARIATE DATA to myopically focus on single gene 
A sin, no less.  Don't kid yourself.

My comment would be: build evidence for biolgoical phenomena ACROSS THE 
10ks of genes
on the array.  Use bioinformatic knowledge to pin together the evidence. 
Let's be SCIENTISTS.  Lets focus on the REAL questions.

All the truths about MTA are still true.  They have use.  FDR is still a 
good idea.
The T distribution is still a pillar.   The question is the appropriate 
scale of biological
inquiry at which to do the testing.  Single gene hit lists just suck. 
 Quote me.


More information about the Bioconductor mailing list