Steve Lianoglou | 13 Sep 19:53 2013

Re: Microarray analyses. How to know the direction of the expression?

Hi,

On Thu, Sep 12, 2013 at 11:57 AM, James W. MacDonald <jmacdon@...> wrote:
> Hi Mary,
>
> This is another reason to use a 'cell means' parameterization rather than a
> 'factor effects' parameterization. This is just statistician blahblah for
> computing a mean for each group rather than a baseline level and a
> difference between the other sample and the baseline.
>
> In your situation, since the second coefficient has WT in the name (and you
> have an intercept term), the intercept is computing the mean of a baseline
> sample (in your case the KO sample), and the GroupWT coefficient is WT - KO.
>
> You could instead do
>
> design <- model.matrix(~0+Group)
> contrast <- matrix(c(1,-1), ncol = 1)
> colnames(contrast) <- "KO vs WT"

Perhaps using a even a little less "statistician blahblah", the OP
might find using the limma::makeContrasts function more intuitive, eg:

R> contrast <- makeContrasts(KO-WT, levels=levels(Group))

And continue as Jim prescribed below:

> fit <- lmFit(myEset, design)
> fit2 <- contrasts.fit(fit, contrast)
> fit2 <- eBayes(fit2)
> topTable(fit2, 1)

Sincerely,
A non-statistician who deals with his fair share of statistical blah blah

--

-- 
Steve Lianoglou
Computational Biologist
Bioinformatics and Computational Biology
Genentech

_______________________________________________
Bioconductor mailing list
Bioconductor@...
https://stat.ethz.ch/mailman/listinfo/bioconductor
Search the archives: http://news.gmane.org/gmane.science.biology.informatics.conductor


Gmane