[BioC] limma advice required. Investigating the amplification of small sample RNA

[Gordon Smyth]

>Date: Wed, 26 Jul 2006 08:20:17 +0100
>From: Andrew Mcdonagh <a.mcdonagh at imperial.ac.uk>
>Subject: [BioC] limma advice required. Investigating the amplification
>         of small sample RNA
>To: bioconductor at stat.math.ethz.ch

[snip]

>  1) How do I carry out the t-test on a per-gene basis given the mean
>protocol effect available from the fit object. I can see that the limma
>guide has a way of obtaining the t-statistics but I'm not really sure
>how to do the test on a per gene basis. I guess it's typical

See the Limma User's Guide section 10.1, or see ?eBayes, or type 
example(eBayes) at the R prompt.

>  2) I look at the stdev.unscaled and it is the same for each protocol. Is
>this to be expected? Sorry, my stats knowledge is not great.
>
>  round_1   round_2   non_amp
>0.7071068 0.7071068 0.7071068

Yes it is to be expected. (Your experiment is symmetric in the three 
protocols, so how could it be otherwise?)

>  3) What is the difference between sigma and stdev.unscaled?

See the limma chapter in the Bioconductor book for a full 
explanation. sigma is just the residual standard deviation.

Best wishes
Gordon

[snip]