[BioC] tilingArray questions

[Wolfgang Huber]

24/10/2008 21:32 Anjan Purkayastha scripsit
> hi,
> a rookie question on tilingArray:
> what criteria does one use to set a value for nrStrata?
> all help will be appreciated.
> tia,
> anjan
> 

Hi Anjan

I assume that you are refering to the parameter "nrStrata" of the
function "normalizeByReference" in the "tilingArray" package.

Good choices of the value are bounded:
- above by the number of background features (if you have too few in a
stratum, the estimation of the background correction becomes unprecise) and
- below by the variability of the values in "refSig" within each
stratum; the background correction assumes that probes within one
stratum have the same (background signal) distribution.

I found "nrStrata=10" a good choice for an Affymetrix array tiling a
yeast gnome with 6.5 Mio probes - you will need to make your own
exploratory analysis of these criteria for your array.
-- 
Best wishes
 Wolfgang

------------------------------------------------------------------
Wolfgang Huber  EBI/EMBL  Cambridge UK  http://www.ebi.ac.uk/huber