Entering edit mode
It is easy to extract ordinary t-statistics from limma, as well as the moderated t-statistics. It is explained in Section 8.1 of the User's Guide -- do a text search for "ordinary.t". The method is
fit <- lmFit( data, design )
ordinary.t <- fit$coef / fit$stdev.unscaled / fit$sigma
Gordon
PS. I have no intention of providing the ordinary t statistic automatically, because it has such poor performance. So if you want it, you have to type one simple line of code.
>michael watson (IAH-C) michael.watson at bbsrc.ac.uk
>Wed Jun 2 15:26:23 CEST 2004
>
>
>Hi
>
>I have a dataset which is pretty much IDENTICAL to the swirl dataset:
>
>Experiment 1 - two replicate arrays with a dye swap:
>
>TreatedCy5 vs UntreatedCy3
>UntreatedCy5 vs TreatedCy3
>
>Experiment 2 - two replicate arrays with a dye swap:
>
>TreatedCy5 vs UntreatedCy3
>UntreatedCy5 vs TreatedCy3
>
>This is fantastic because I can basically just copy and paste the
>instructions from the limma userguide.pdf document to get my
>differentially expressed genes.
>
>However, I want to do a comparison of limma with a "normal" method of
>analysis - say a t-test. How would I carry out a t-test on this kind of
>data?
>
>For example, the top gene limma pulls out has these values for my four
>arrays for log2(Cy5/Cy3):
>
>-2.7, 2.7, -2.7, 3
>
>This makes sense as the experiment contains a dye-swap, so if I flip my
>log(ratios) such that I am always comparing treated/untreated, my values
>are -2.7, -2.7, -2.7 and -3. BUT how would I go about doing a t-test on
>this kind of comparison??? (I know there are huge arguments against
>doing such a thing, but humour me). I mean, I basically have four
>values for the same thing (the relative expression of treated against
>untreated) and if I was doing a t-test - what am I comparing the
values against?
>
>Thanks in advance
>
>Mick
