[BioC] Limma and kooperberg background correction

[michael watson (IAH-C)]

Hi

I have noticed while comparing kooperberg background correction with
normal background correction that on my M vs A plots, the kooperberg
method seems to have a LOT more negative M values at low A values than
the normal background correction method - in fact, it makes the M vs A
plots look more like "hump backs" than the traditional "nike swoosh" (if
anyone can tell what I am talking about!)

Can anyone think why kooperberg might create more negative log ratios??
I realise that most of the effect of kooperberg will be at the low-A end
of the spectrum, but why would it necessarily create more negative
ratios than postive ratios?  Perhaps my negative values are only in one
channel....??

Thanks

Mick