[BioC] Explaining limma, design matrices and linear models to non statisticians
Natalie P. Thorne
npt22 at cam.ac.uk
Mon Feb 27 16:31:20 CET 2006
I have some slides on how to make design matrices that I use to teach
biologists. Actually I've been thinking, for a while, about posting them
somewhere on Bioconductor for general resource.
I have several slides for two-colour data and a couple for single-color
data. The two-color examples, start with very basic experimental designs,
two samples only, through to 3 samples (direct and ref designs), 4 samples
factorial design and 2x3 factorial/time series designs.
The slides work through how to select the parameters, including
alternative parameterisations, how to interpret them, and how to specify
them in a design matrix. Its very hands on - I usually get people to work
through each example by hand so they really get to understand the idea of
matrix multiplication of the design matrix with the parameter vector and
how it actually relates to what is measured on each slide.
I've used these over a dozen times for various courses and I find they
work quite well...of course I usually modify and add slightly to them each
time. You are welcome to try them if you like.
BTW: I also have 4 practicals that I use to teach the fundamentals of
limma (they include RBasics, LimmaBasics, LimmaPreprocessing,
LimmaDiffExpression). These have a lot more detail for real beginners
to R who are trying to use Limma. The examples are complementary to
the LimmaUsers guide, but they include more detail on the basics that
biologists often find hard when they're getting started. The problem is,
the latter three pracs come with data sets that will probably exceed the
space limit acceptable by BioConductor. Also I have the beginnings of a
Limma "Reference Card". I guess, I should figure out a sensible way/place
to post all this.
On Mon, 27 Feb 2006, michael watson (IAH-C) wrote:
> I'm going to be teaching use of limma to a bunch of non-statisticians
> later this month, and I wondered if there were any slides or other
> resources that I could possibly hi-jack to explain how and why we create
> design matrices?
> Many thanks
Natalie Thorne, PhD
Computational Biology Group
Hutchison/MRC Research Centre
Department of Oncology
University of Cambridge
Hills Rd, Cambridge CB2 2XZ
Email: npt22 at cam.ac.uk
Phone: +44 (0)1223 763381
Fax : +44 (0)1223 763262
More information about the Bioconductor