[BioC] Limma -Time series data

[James W. MacDonald]

kuangqin wrote:
> Dear James,
> Thank you very much for your reply!
>  
> For my question:
>> Which genes have estrogen effect (up/down between absent and present) > CONSISTENTLY at both 10 hours and 48 hours?> 
>  I am more interested in whether I can  add one contrast in either design or contrast matrix to dothis job ???
>  
> Is it the one way to look the intersection of the two comparisons for answering the question?

It depends on what you mean by 'consistently'. If you mean they both go 
the same direction and are significant, then decideTests() is the way to 
go.

If you mean they both go the same direction and have very similar 
magnitudes, then you could look for genes that are significant in both 
E10 and E48 (and with the same sign) but not in the interaction.

You could also add a column to the contrasts matrix (c(0,0,0.5,0.5)), 
but that tests to see if the average of coefficients E10 and E48 is 
significant, so if one coefficient is large and the other small, you can 
still get significance even though one is not strictly significant.

-- 
James W. MacDonald, M.S.
Biostatistician
Affymetrix and cDNA Microarray Core
University of Michigan Cancer Center
1500 E. Medical Center Drive
7410 CCGC
Ann Arbor MI 48109
734-647-5623