[BioC] Rosetta vs Limma

[Gordon K Smyth]

> Date: Thu, 17 Nov 2005 21:55:44 +0100
> From: Nataliya Yeremenko <eremenko at science.uva.nl>
> Subject: [BioC] Rosetta vs Limma
> To: Bioconductor List <bioconductor at stat.math.ethz.ch>
>
> Hello
> I have a questions about correlation of outcomes from Rosetta and Limma.
> I have done Agilent direct hyb's (one condition versus another).
> In Rosetta I have splitted ratios without common references and than
> have done two sample Ttest.

I'm not sure what you mean by "splitted ratios".  I am guessing that you have done a two-sample
t-test using single channel log-ratios.  This is a very poor method, so I'm glad that limma
doesn't give exactly the same results.

Why do you find it surprising that different methods of analysis give somewhat different results? 
You describe it as a "problem".  Why is it a surprise that good analysis methods really do make a
difference?

Best wishes
Gordon

> In Limma I also splitted channels and than fitted linear model. I have
> posted design in previous BioC thread with subj "Design question in LIMMA".
> The outcome of the Toptable is different (from 120 signatures only 20
> intersects).
> Does anybody experienced that problem or have expertise of statistical
> treatment of twocolors direct ratio based arrays
> in Rosetta and Limma simultaneously.
>
> Regards,
> Nataliya
>
>
>
> --
> Dr. Nataliya Yeremenko
>
> Universiteit van Amsterdam
> Faculty of Science
> IBED/AMB (Aquatische Microbiologie)
> Nieuwe Achtergracht 127
> NL-1018WS Amsterdam
> the Netherlands
>
> tel. + 31 20 5257089
> fax  + 31 20 5257064