[BioC] Bayesian selection of genes and its correct template

[Gordon Smyth]

At 09:45 PM 16/07/2004, Ian Jeffery wrote:
>hi,
>   I was looking at the Bayesian method of selecting genes. I used the
>commands
>
>library(limma)
>fit <- lmFit(x,design=y)#x is the datafile
>                         #y is the template
>eb2<- eBayes(fit)
>topTable(eb2,coef=,number=15,adjust="fdr")
>
>I found that that the two groups could be defined by the template in two
>ways
>either
>       (0,0,0,0,1,1,1,1)
>or
>       (-1,-1,-1,-1,1,1,1,1)
>(I tryed other templates and all give one of the two gene lists)
>Both methods return a gene list, so my questions are;
>1. Are both methods correct?

You don't tell us anything about your experiment, so I am assuming that you 
have Affymetrix data with 8 chips in two groups. In that case both of your 
"templates" are wrong. Please read the limma User's Guide. If on the other 
hand you have two colour data with the second 4 arrays a dye-swap of the 
first 4, then your second "template" is correct. The first "template" could 
not be correct for any experiment.

Gordon

>2. If so why do they return different gene lists and what is the
>significance of this?
>
>thanks, Ian