[BioC] robustspline method for within arrays normalisation

[Gordon Smyth]

At 11:37 PM 15/03/2007, Muller, Pie wrote:
>Dear Gordon,
>
>Apologies should I have missed that this topic 
>has already been discussed in the forum. We are 
>using a small scale, custom made array 
>containing 1760 features (440 probes in four 
>replicates). Currently I am using the "loess" 
>function in normalizeWithinArrays() for 
>normalisation. I was wondering whether perhaps 
>the "robustspline" function would be more 
>appropriate but I am somewhat unclear what the 
>differences between the two methods are and 
>whether it would be suitable at all for our small scale arrays.

robustspline is designed to be intermediate 
between "printtiploess" and "loess", depending on 
the data. It's not particularly aimed at small 
arrays, although it can cope with small print-tip 
groups in a way that printtiploess can't.

The greater issue with with custom made arrays is 
whether your probes have been specially chosen so 
that more than half can be differentially 
expressed in a given situation. This issue is explored in

   http://genomebiology.com/2007/8/1/R2

One could argue that robustspline is a little 
more resistant to this than loess, but there's not much in it, see

   https://www.stat.math.ethz.ch/pipermail/bioconductor/2005-November/011006.html

>  Is there any more information available on 
> that method other than already in the Limma user's guide?

No, the limma document is all that's available.

Best wishes
Gordon

>Thanks,
>Pie
>
>-------------------------------------
>
>Dr Pie Müller
>Vector Group
>Liverpool School of Tropical Medicine
>Pembroke Place
>Liverpool
>L3 5QA
>UK
>
>Tel +44(0) 151 705 3225
>Fax +44(0) 151 705 3369
>http://www.liv.ac.uk/lstm