[BioC] linear model analysis

[Elizabeth Brooke-Powell]

Good Afternoon,
 
I have been analyzing something like 150 arrays worth of data using
LimmaGUI, but one thing I am finding as a biologist hard to get a grip of is
the comparability of the B statistic numbers generated. Is a value of 6 for
a gene in one experiment directly comparable to a value of 10 for the same
gene in a different experiment? One other question I had was on a log odds
plot is the log fold change a log based 10 measures so that a change of log
fold change 1 would be 10 in reality? Is this the same as the M value given
in the top table lists? 
 
We are also working with some more complex experimental designs, and I was
wondering if anyone had a paper explaining how the linear model is used for
different experimental designs in particular a saturated design. Does the
model utilize each channel independently taking into account that they are
related? In other words will data be taking from only the arrays with direct
comparison for the treatments of interest or does it take all of the related
labelled sample data independent of the on array comparison? In which case,
do you think is reasonable to add arrays used with different designs for
analysis that contain the same RNA used for labelling? 
 
One final question I have for you all is can the linear model fitting be
used with samples/treatments that seem to have global gene regulation
effects?
 
Thank you all for your help in advance. Hope these are not ridiculous
questions.
 
Yours Sincerely,
 
Liz Brooke-Powell
 
Molteno Building
Department of Pathology
University of Cambridge
Tennis Court Road mbridge, CB2 1QP
United Kingdom
 
Website: http://www.path.cam.ac.uk/~toxo/
Tel 01223 33 33 31(office) or 01223 33 33 29 (lab)
 

	[[alternative HTML version deleted]]